Ameliorative Effects of a Naphthoquinone Derivative With β‐Amyloid Aggregation Inhibitory Activity on Cognitive Impairment and Metabolite Analysis of the Blood and Brains of Mice

ABSTRACT

Accumulation of amyloid‐β (Aβ) plaques is an important cause of Alzheimer's disease (AD) pathogenesis. In this study, we evaluated Aβ aggregation inhibitory activity of synthesized naphthoquinone derivatives as well as improvement in cognitive functions and metabolite profiling of brain tissues using scopolamine (SCO)‐induced mice. Compound 888 (2‐(4‐(2,3,4‐trimethoxybenzyl)piperazin‐1‐yl)naphthalene‐1,4‐dione, [TPN]) showed the highest Aβ aggregation inhibitory activity (IC₅₀ = 0.14 μM), and was more potent than the reference compound curcumin (IC₅₀ = 1.63 μM). Compound TPN showed effective monoamine oxidase (MAO)‐A, MAO‐B, acetylcholinesterase, and butyrylcholinesterase inhibitions at 10 μM, likely as candidates for multitarget‐directed ligands. TPN was permeable through the blood‐brain barrier, and non‐toxic to MDCK and SH‐SY5Y cells. TPN displayed prolonged and stable interactions with Aβ₄₂ during molecular dynamics simulations, in contrast to the short‐lived contacts observed for curcumin. Cognitive impairment was significantly improved by TPN‐treatment in behavioral tests. TPN treatment attenuated Aβ‐related protein expression, inflammatory responses, oxidative stress‐related changes, and apoptosis‐related alterations, while preserving hippocampal pyramidal neurons and their typical morphology. In metabolite profiling, TPN modulated a narrower set of pathways mainly related to amino acid and kynurenine metabolism, whereas donepezil induced broader adjustments involving amino acid, mitochondrial/energy, and lipid‐related pathways compared to those in the serum and cortex of the SCO group, in contrast to those in the hippocampus. Collectively, a potent Aβ aggregation inhibitor TPN showed significant cognitive improvement, accompanying by neuroprotective effects, decreasing inflammation, and retaining neuron structures, exhibiting changed metabolic profiles compared to the control treatments. These findings suggest that TPN has cognitive‐protective and neuroprotective potential under scopolamine‐induced impairment conditions and warrants further validation in AD‐relevant models.

1. Introduction

Alzheimer's disease (AD), the leading cause of dementia, is one of the major causes of death in the world. The prevalence of AD continues to increase over time, underscoring the growing importance of research into neurodegenerative disorders and the development of effective therapeutic agents. The primary symptoms of AD are cognitive impairment and behavioral disturbances, and currently approved therapeutic drugs provide only short‐term symptomatic relief. Various studies using AD model types have shown that AD is characterized by the accumulation of amyloid‐β (Aβ) plaques, tau hyperphosphorylation, and neuronal loss, which occur over a prolonged period of approximately 20–30 years (Dubois et al. 2014; Hampel et al. 2021). In addition to Aβ‐mediated plaque deposition, other hallmark histopathological features of AD include intracellular neurofibrillary tangles and synaptic degeneration (Butterfield and Boyd‐Kimball 2004; Ittner et al. 2010). These alterations occur predominantly in the neocortex, hippocampus, and other subcortical regions critical for cognitive function (Nordberg 2008). As these pathological changes emerge prior to the onset of clinical symptoms, they may serve as valuable biomarkers for the early diagnosis of AD.

Amyloid precursor protein (APP), a single‐pass transmembrane protein primarily expressed in the brain, undergoes proteolytic processing through two distinct pathways mediated by α‐ or β‐secretases (especially β‐site APP cleaving enzyme 1 [BACE1]). In the non‐amyloidogenic pathway, APP is cleaved by α‐ and subsequently γ‐secretases producing soluble APP with a truncated amino acid sequence that is non‐neurotoxic (Patterson et al. 2008). In contrast, in the amyloidogenic pathway, sequential cleavage by β‐ and γ‐secretases generates Aβ peptides consisting of 38–43 amino acids. Among them, Aβ₄₂ and Aβ₄₃ exhibit particularly high aggregation propensity, existing as monomers, oligomers, and protofibrils, as well as protofibrils that form fibrils and aggregate as plaques, thereby contributing to neurotoxicity in the brain (Patterson et al. 2008; Harkany et al. 2000; Selkoe 1998; Zhao et al. 2020). In addition, Aβ aggregates interact with tau to induce toxicity, while phosphorylated tau further promotes Aβ‐induced mitochondrial damage in neurons (Do et al. 2014). Recently, aducanumab (Sevigny et al. 2016), lecanemab (Van Dyck et al. 2023), and donanemab (Mintun et al. 2021) have been developed as therapeutic agents for AD by inhibiting Aβ aggregation. However, adverse effects associated with these drugs have been reported (Salloway et al. 2022; Walsh et al. 2024; Zimmer et al. 2025), highlighting the need to develop new treatment agents. The scopolamine‐induced cognitive impairment model is a widely used pharmacological model for evaluating learning and memory deficits associated with cholinergic dysfunction. Because cholinergic impairment is closely related to cognitive decline, this model provides a practical in vivo platform for the preliminary evaluation of compounds with cognitive‐enhancing and neuroprotective potential. In the present study, this model was employed to examine whether TPN could ameliorate scopolamine‐induced behavioral deficits and associated molecular and histological alterations in mouse brain tissues.

Quinone and naphthoquinone scaffolds are highly attractive structures in medicinal chemistry that are consistently isolated from natural sources and renowned for their diverse and potent biological profiles. Naphthoquinone derivatives have been reported to exhibit biological activities, such as anticancer (Shao et al. 2023; Zhu et al. 2022), antibiotic (Song et al. 2020; Yap et al. 2021), anti‐inflammation (Kobayashi et al. 2011). In addition to their well‐established roles as anticancer and antimicrobial agents, previous studies have highlighted their strong antioxidant and neuroprotective properties. Specifically, relevant to our recent studies, key naphthoquinone derivatives such as plumbagin (Nakhate et al. 2018; Anand et al. 2021) and β‐lapachone (Mokarizadeh et al. 2020; Lee et al. 2025, 2015) were shown to have significant potential to address multiple AD‐related pathologies. These molecules are known to provide neuroprotection not only by mitigating oxidative stress and modulating neuroinflammation but, critically, by directly inhibiting Aβ aggregation.

Recently, due to the failure of limited therapeutic approaches caused by the diverse etiologies and mechanisms of AD, research and development have focused not only on single target inhibitors but also on multitarget inhibitors as multitarget‐directed ligands (MTDLs) that can inhibit two or more targets such as dual target inhibitors acting on Aβ aggregation and cholinesterase (ChE) or on Aβ and monoamine oxidases (MAOs) (Cacabelos et al. 2024; Wang et al. 2019; Pan et al. 2014). The 1,4‐naphthoquinone scaffold has emerged as a versatile and privileged framework to develop MTDLs for AD therapy (Campora et al. 2021a, 2021b). This pharmacological relevance stems from its intrinsic ability to simultaneously modulate acetylcholinesterase (AChE) activity, inhibit Aβ aggregation, and counteract oxidative stress—three key pathological processes implicated in AD progression (Campora et al. 2021a, 2021b; Khelifi et al. 2020). Notably, structure‐activity relationship (SAR) analyses of naphthoquinone‐based hybrids have revealed that their planar aromatic cores enable productive interactions with amyloidogenic peptides, leading to disruption of π–π stacking interactions essential for fibril formation (Paul et al. 2019).

Building upon the inherent bioactivity of the 1,4‐naphthoquinone scaffold and prior evidence supporting the efficacy of naphthoquinone‐based hybrids in antagonizing amyloid aggregation (Paul et al. 2019), we designed and synthesized a novel hybrid molecule: 2‐(4‐(2,3,4‐trimethoxybenzyl)piperazin‐1‐yl)naphthalene‐1,4‐dione (compound 888, abbreviated as TPN). This design represents a deliberate departure from traditional naphthoquinone analogues that rely exclusively on standard planar aromatic cores to inhibit amyloidogenic π–π stacking interactions (Paul et al. 2019). By strategically grafting a trimetazidine‐derived 2,3,4‐trimethoxybenzylpiperazine moiety onto the C2 position of the core ring, we introduced a distinct architectural component that fundamentally expands the scaffold's binding topology and multi‐functional capability.

This pharmacophore is a defining structural feature of trimetazidine, a clinically established cytoprotective agent widely used for the management of ischemic heart disease (Chrusciel et al. 2014). Trimetazidine (1‐[(2,3,4‐trimethoxyphenyl)methyl]piperazine) exerts its protective effects primarily through metabolic modulation—selectively inhibiting fatty acid β‐oxidation and promoting glucose oxidation, which improves cellular energetic efficiency under ischemic conditions and, consequently, attenuates oxidative stress‐related damage (Chrusciel et al. 2014). The rationale for this molecular hybridization is to integrate the established anti‐amyloidogenic and redox‐modulating properties of the 1,4‐naphthoquinone scaffold—specifically its capacity to interfere with Aβ aggregation and alleviate oxidative stress (Khelifi et al. 2020; Paul et al. 2019)—with the metabolic modulatory effects of trimetazidine‐derived pharmacophores, which improve cellular energetic efficiency by shifting substrate utilization toward glucose oxidation (Chrusciel et al. 2014). Mechanistically, this structural innovation offers two distinct advantages over existing literature: (i) the substituted piperazine ring imparts tailored spatial flexibility that optimizes steric interaction within the amyloid assemblies, enabling strong, prolonged contacts with crucial regulatory residues such as ASP7, GLN15, and GLU22 that govern Aβ₄₂ aggregation kinetics, and (ii) it systematically addresses the early‐stage mitochondrial energy crises and metabolic failures that render neuronal circuitries vulnerable in the AD brain (Kapogiannis and Mattson 2011). This structural evolution thus transitions the naphthoquinone core from a standard single‐target anti‐aggregant into a specialized metabolic–neuroprotective multi‐target lead, clear distinguishing its therapeutic mechanism from conventional analogues.

On the other hand, AD, traditionally defined by Aβ aggregation and tau pathology, is increasingly recognized as a condition accompanied by metabolic disturbances that extend across both the central nervous system (CNS) and peripheral tissues. Recent applications of metabolomics have provided detailed biochemical characterizations of these alterations, enabling the detection of endogenous metabolite changes associated with AD progression. Multiple studies have reported consistent perturbations in amino acid metabolism, lipid remodeling, mitochondrial energy pathways, and redox‐related processes in the plasma, cerebrospinal fluid, and postmortem brain tissue from patients with AD and in transgenic AD mouse models (Trushina et al. 2013; Paglia et al. 2016; Holubiec et al. 2022; Zhou et al. 2023; Liu et al. 2022). Several of these metabolic abnormalities have also been observed during the early or prodromal stages, suggesting their potential utility as indicators of emerging neuropathology (Franco et al. 2024; Lista et al. 2023).

Metabolomics has been employed to characterize metabolic responses to therapeutic interventions. Compounds with anti‐amyloid, antioxidant, and neuroprotective activities have been associated with measurable shifts in metabolic pathways, including aminoacid turnover, mitochondrial‐related metabolites, and lipid mediators (Yin 2023; Bilski et al. 2025). These treatment‐associated metabolic changes provide important complementary information beyond histopathological assessments and support the use of metabolomics to monitor biochemical responses following drug administration. Therefore, evaluating metabolite profiles in blood and brain tissues after exposure to candidate compounds, such as naphthoquinone derivatives, may offer insight into how these agents influence central and systemic metabolic networks. Additionally, metabolites that consistently respond to treatment may serve as informative candidates for monitoring the protective effect or biochemical changes during intervention studies (Apiraksattayakul et al. 2024).

The objective of this study is a modification of the naphthoquinone scaffold to maximize its affinity and inhibitory potency against Aβ aggregation, while concurrently enhancing its neuroprotective characteristics. Further, this study aims to introduce novel, synthetically accessible naphthoquinone compounds as promising ameliorative agents against scopolamine‐induced cognitive impairment in mice, which can be used for robust candidates for disease‐modifying AD therapy. In this study, we synthesized the novel naphthoquinone derivatives and evaluated their potential as Aβ aggregation inhibitors, as well as molecular docking and cognitive improvement analyses with a lead compound by animal behavioral tests as well as Western blotting analysis for key protein expression levels, and further performed metabolite profiling for the blood and brain tissues.

2. Materials and Methods

2.1. Chemistry

2.1.1. Synthesis of 587

At first, 100 mg of 1,4‐dihydroxy naphthalene (0.62 mmol) was dissolved in 10 mL of EtOH. Next, 20 mg of CeCl₃·7H₂O (0.062 mmol), 120 mg of tryptamine (0.75 mmol), and 130 mg of TEA (1.25 mmol) were sequentially added dropwise. The mixture was then stirred at room temperature for 6 h. After confirming completion of the reaction by thin layer chromatography (TLC) (DCM:MeOH = 19:1), the resulting solid was collected by filtration after recrystallization from water to acquire compound 587 (82% yield). Melting Point (MP): 178°C. ¹H NMR (DMSO‐d ₆, 400 MHz) δ 3.01 (t, J = 7.2 Hz, 2H), 3.48 (m, 2H), 5.74 (s, 1H), 6.98 (t, J = 7.2 Hz, 1H), 7.07 (t, J = 7.6 Hz, 1H), 7.26 (d, J = 1.5 Hz, 1H), 7.34 (d, J = 8 Hz, 1H), 7.52 (t, J = 5.8 Hz, 1H), 7.57 (d, J = 7.8 Hz, 1H), 7.72 (t, J = 7 Hz, 1H), 7.82 (t, J = 7 Hz, 1H), 7.96 (m, 2H), 10.86 (s, 1H); ¹³C NMR (DMSO‐ d ₆, 100 MHz) δ 23.81 (1 C), 43.11 (1 C), 99.86 (1 C), 111.65 (1 C), 111.91 (1 C), 118.67 (1 C), 118.82 (1 C), 121.49 (1 C), 123.56 (1 C), 125.83 (1 C), 126.38 (1 C), 127.61 (1 C), 130.85 (1 C), 132.63(1 C), 133.69 (1 C), 135.35(1 C), 136.73 (1 C), 148.84 (1 C), 181.74 (1 C), 182.05 (1 C); ESI–MS: m/z [M + H]⁺ 317.1 (calcd. 316.12). SMILES: O═C1C2═CC═CC═C2C(C═C1NCCC3═CNC4═C3C═CC═C4)═O.

2.1.2. Synthesis of 591

Compound 591 was synthesized following the procedure described for 587 by substituting tryptamine with 140 mg of N,N‐dibutylaminopropylamine (0.75 mmol). The target compound was obtained in a 38% yield. ¹H NMR (DMSO‐d ₆, 400 MHz) δ 0.85 (t, J = 7.2 Hz, 6H), 1.28 (s, J = 7.2 Hz, 4H), 1.37 (q, J = 7.2 Hz, 4H), 1.70 (q, J = 6.4 Hz, 2H), 2.34 (t, J = 7.2 Hz, 4H), 2.44 (t, J = 6.4 Hz, 2H), 3.20 (q, J = 6.0 Hz, 2H), 5.64 (s, 1H), 7.72 (t, J = 7.6 Hz, 1H), 7.80 (t, J = 6.0 Hz, 1H), 7.82 (t, J = 7.6 Hz, 1H), 7.94 (d, J = 7.6 Hz, 1H), 7.98 (d, J = 7.6 Hz, 1H); ¹³C NMR (DMSO‐ d ₆, 100 MHz) δ 14.188 (2 C), 20.608 (2 C), 25.210 (1 C), 29.305 (2 C), 41.568 (1 C), 52.255 (1 C), 53.778 (2 C), 99.643 (2 C), 125.808 (1 C), 126.329 (1 C), 130.854 (1 C), 132.567 (1 C), 135.290 (1 C), 149.033 (1 C), 181.605 (1 C), 182.023 (1 C); ESI–MS: m/z [M + H]⁺ 343.3 (calcd. 342.23). SMILES: O═C1C(NCCCN(CCCC)CCCC)═CC(C2═CC═CC═C21)═O.

2.1.3. Synthesis of 671

Compound 671 was synthesized following the procedure described for 587 using 101.9 mg of N,N‐dimethyl‐p‐phenylenediamine (1.2 eq.) instead of tryptamine. The solid product was obtained with an 80% yield. Melting Point (MP): 180°C. ¹H NMR (DMSO‐d ₆, 400 MHz) δ 2.92 (s, 6H), 5.92 (s, 1H), 6.79 (d, J = 8.8 Hz, 3H), 7.2 (d, J = 8.4 Hz, 2H), 7.76 (t, J = 7.2 Hz, 1H), 7.85 (t, J = 7.2 Hz, 1H), 7.94 (d, J = 7.2 Hz, 1H), 8.05 (d, J = 7.6 Hz, 1H), 9.09(s, 1H); ¹³C NMR (CDCl₃, 100 MHz) δ 40.634 (2 C), 102.013 (1 C), 113.029 (2 C), 124.531 (2 C), 126.109 (2 C), 126.378 (1 C), 130.537 (1 C), 132.005 (1 C), 133.661 (1 C), 134.799 (1 C), 145.756 (1 C), 148.826 (1 C), 182.391 (1 C), 183.551 (1 C); ESI–MS: m/z [M + H]⁺ 293.28 (calcd. 292.12). SMILES: O═C1C(NC2═CC═C(N(C)C)C═C2)═CC(C3═CC═CC═C31)═O.

2.1.4. Synthesis of 888 ((2‐(4‐(2,3,4‐trimethoxybenzyl)piperazin‐1‐yl)naphthalene‐1,4‐dione, TPN)

Compound TPN was prepared following the procedure described for 587 by substituting tryptamine with 410 mg of 2,3,4‐trimethoxybenzyl piperazine (1.20 mmol). The target compound was obtained with a 35% yield. ¹H NMR (CDCl₃, 400 MHz) δ 2.63(t, J = 4.8 Hz, 4H), 3.51 ~ 3.53 (m, 6H), 3.86 (s, 3H), 3.88 (s, 3H), 3.90 (s, 3H), 6.00 (s, 1H), 6.65(d, J = 8.8 Hz, 1H), 6.99(d, J = 8.4 Hz, 1H), 7.62 (td, J = 1.2 Hz, J = 7.6 Hz, 1H), 7.63 (td, J = 1.2 Hz, J = 7.6 Hz, 1H), 7.99 (dd, J = 0.8 Hz, J = 7.6 Hz, 1H), 8.03 (dd, J = 0.8 Hz, J = 7.6 Hz, 1H); ¹³C NMR (DMSO‐ d ₆, 100 MHz) δ 49.03 (2 C), 52.42 (2 C), 55.99 (1 C), 56.43 (1 C), 60.80 (1 C), 61.22 (1 C), 107.00 (1 C), 111.47 (1 C), 123.30 (1 C), 125.17 (1 C), 125.52 (1 C), 126.65 (1 C), 132.40 (1 C), 132.84 (2 C), 133.83 (1 C), 142.34 (1 C), 152.67 (1 C), 153.12 (1 C), 153.80 (1 C), 183.11 (1 C), 183.64 (1 C); ESI–MS: m/z [M + H]⁺ 423.2 (calcd. 422.18). SMILES: O═C1C(N2CCN(CC3═C(OC)C(OC)═C(OC)C═C3)CC2)═CC(C4═C1C═CC═C4)═O.

2.2. Aβ Aggregation and Enzyme Inhibition Assays

The β‐amyloid aggregation inhibitory activity was evaluated using Thioflavine T (ThT) assay method as previously described (Oh et al. 2025a), with slight modifications. For the assay, Aβ(1‐42) (Aβ₄₂) used was hexafluoro‐2‐propanol (HFIP)‐treated peptide (GenicBio Ltd., Shanghai, China). The HFIP‐Aβ₄₂ was dissolved in 50 mM Tris (pH 7.2) containing 150 mM NaCl to 0.25 mM with 3 min sonication and final 60 μM was used for standard assay in 100 μL. After addition of 0.5 mM ThT, the fluorescence was measured using a microplate spectrophotometer (Varioskan LUX, Thermo Fisher Scientific Inc.) in a 96‐well microplate for 3 h at 37°C with excitation and emission wavelengths at 440 and 484 nm, respectively. BACE1 inhibitory activity was evaluated using a BACE1 assay kit (CS0010, Sigma‐Aldrich) in a 96‐well plate and the microplate spectrophotometer at excitation and emission wavelengths of 320 and 405 nm, respectively (Oh et al. 2025b). Human MAO‐A (recombinant, M7316, Sigma‐Aldrich) and MAO‐B (recombinant, M7441, Sigma‐Aldrich) activities were measured using a continuous assay protocol at 316 and 250 nm for 45 min (Oh et al. 2021, 2023). AChE from Electrophorus electricus (C2888, Sigma‐Aldrich) and butyrylcholinesterase (BChE) from quine serum (C1057, Sigma‐Aldrich) inhibitory activity evaluated using a continuous assay with 5,5‘‐dithiobis (2‐nitrobenzoic acid) (DTNB) with acetylthiocholine iodide (ATCI) and S‐butyrylthiocholine iodide (BTCI) as substrate, and the absorbance was measured at 412 nm (Oh et al. 2023; Heo et al. 2020).

2.3. Molecular Docking and Molecular Dynamics Simulation

Molecules, including TPN and curcumin were docked to Aβ₄₂ using AutoDock Vina (Eberhardt et al. 2021). Model 1 of the Aβ NMR structure was used as the receptor structure (PDB ID: 1Z0Q) (Tomaselli et al. 2006). A docking box covering the entire structure was constructed. Three‐dimensional (3D) structures of the ligands were generated using an employing RDKit (Landrum 2026). The atomic partial charges and protonation states of the receptor and ligands were assigned by Meeko (Santos‐Martins et al. 2025). The interactions between proteins and ligands were analyzed using PLIP (Schake et al. 2025).

A molecular dynamics (MD) study was conducted on the complex structure of Aβ and three molecules: TPN, 671, and curcumin. A periodic boundary condition was applied to the cubic box, generated by extending 10 Å along each axis from each end of the complex. The complex structure was solvated with TIP3P water and neutralized by adding potassium and chlorine ions at a concentration of 0.15 M. For the protein and the ligand, the Charmm36m force field and the CGenFF parametrization were applied, respectively. Input preparation was executed using CHARMM‐GUI (Lee et al. 2016). GROMACS 2024.4 (Pronk et al. 2013) was used to perform the MD simulations. The solvated system was subjected to 5000 steps of steepest‐descent minimization. After equilibration of 500 ps, 200 ns of MD production was performed at 303.15 K with a van der Waals cutoff of 12 Å. The LINCS was applied to ensure a time step of 2 fs.

2.4. In Silico Pharmacokinetic Properties

The pharmacokinetics of broussochalcone A were predicted for gastrointestinal (GI) absorption, brain‐blood barrier (BBB) permeability, P‐glycoprotein (P‐gp) substrate, cytochrome P450 inhibition, and skin permeation using the SwissADME web tool (http://www.Swissadme.ch/, accessed on November 2, 2025) (Daina et al. 2017).

2.5. Cell Study

2.5.1. Cell Culture

Human neuroblastoma SH‐SY5Y and Madin‐Darby canine kidney (MDCK) cells were provided by the Korean Cell Line Bank (KCLB). The cells were cultured in minimum essential medium (MEM; Gibco, USA) and Dulbecco's Modified Eagle's Medium (DMEM) with 10% fetal bovine serum (FBS; Gibco, USA) and 1% penicillin streptomycin. The cells cultured in 37°C and 5% CO₂ atmosphere.

2.5.2. Cytotoxicity and neuroprotective activity

SH‐SY5Y and MDCK cells were seeded at density of 5 × 10⁴ cells/well in 96‐well plates. After 24 h, the cultured cells were treated with 671 or TPN for an additional 24 h. For the cell study section, compound TPN was referred to as TPN. Cytotoxicity was evaluated using the cell counting kit‐8 (CCK‐8, Dojindo, Japan) and a microplate reader (Multiskan FC Microplate Photometer, Thermo Fisher Scientific). Cell viability was determined by measuring absorbance at 450 nm (Oh et al. 2023). To evaluate the neuroprotective effect, Aβ₄₂ (GenicBio Ltd., Shanghai, China) was dissolved in DMSO to a final concentration of 10 μM, and then SH‐SY5Y cells were co‐treated with Aβ₄₂ and TPN, the leading compound, at the same time, and cell viability was measured after 24 h.

2.6. Animal Behavioral Experiments

All animal experiments were conducted in accordance with the principles of laboratory animal care (NIH Publication No. 85‐23, revised 1996) and institutional guidelines approved by the Institutional Animal Care and Use Committee of the Korea Institute of Oriental Medicine (KIOM‐IACUC, approval No. 25‐021). Seven‐week‐old male C57BL/6 mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). Animals were randomly assigned to groups (n = 8 per group) and maintained under specific pathogen‐free conditions at 23 ± 1°C with a 12 h light/dark cycle, with ad libitum access to standard feed (Purina Korea, Seoul, Korea) and water. Scopolamine (SCO) was administered intraperitoneally at 1 mg/kg/day from day 14 to day 35 to induce cognitive impairment. TPN was administered intraperitoneally at 10 mg/kg/day from day 1 to day 35, beginning before SCO exposure and continuing throughout the SCO treatment period to evaluate preventive and therapeutic effects. Donepezil (DPZ), used as a positive control, was administered orally at 5 mg/kg/day from day 14 to day 35. Normal control animals received physiological saline according to the same experimental schedule. Behavioral tests were performed during the final phase of the experiment according to the schedule shown in Figure 1.

Figure 1.

Time schedule of the animal behavioral study for the drug TPN (compound 888 ((2‐(4‐(2,3,4‐trimethoxybenzyl)piperazin‐1‐yl)naphthalene‐1,4‐dione) using mice (n = 8). After 7 days of acclimatization, TPN (10 mg/kg/day, i.p.) was administered from day 1 to day 35. Scopolamine (SCO, 1 mg/kg/day, i.p.) and donepezil (DPZ, 5 mg/kg/day, p.o.) were administered from day 14 to day 35. Behavioral tests, including the Morris water maze, Y‐maze, and passive avoidance tests, were performed during the final phase of the experiment as indicated in the timeline. NOR, normal control group; SCO, scopolamine‐treated group; TPN, scopolamine + TPN‐treated group; DPZ, scopolamine + donepezil‐treated positive control group.

The Morris Water Maze (MWM) test was used to evaluate spatial learning and memory of the mice. The maze consisted of a circular rubber pool (diameter: 1.2 m) filled with opaque water (diluted with white dye) and a hidden platform (diameter: 10 cm) in the northeastern quadrant. Swimming activity was recorded and analyzed using an overhead video camera connected to the SMART video tracking software (SMART v3.0, Panlab SL, Barcelona, Spain). On day 0, the mice underwent habituation training consisting of a 60 s session once daily for four consecutive days, during which the platform was visible (1 cm above the water surface) and the water was clear. Test trials were conducted over five consecutive days (days 1–5). All behavioral procedures were performed as previously described (Vorhees and Williams 2006; Bae et al. 2024).

The Y‐maze apparatus was constructed using black polyvinyl plastic and consisted of three arms (42 cm long, 3 cm wide, and 12 cm high) arranged at an angle of 120°. The arms were randomly designated as A, B, or C. Each mouse was placed at the end of one arm and allowed to freely explore the maze for 8 min. Arm entries were recorded, and a valid entry was defined as the placement of all four paws within an arm. Re‐entries into the same arm were also recorded. Spontaneous alternation behavior was defined as consecutive entries into all three arms in overlapping triplet sets (e.g., ABC, BCA, or CAB).

The passive avoidance test (PAT) was performed using a Shuttle Box Avoidance Basic Test Package (Med Associates Inc., Fairfax, VT, USA). The apparatus consisted of two compartments, a light chamber and a dark chamber, separated by an automated sliding door. The floor of the dark chamber was equipped with an electrifiable grid to deliver foot shocks. Step‐through latency, defined as the time taken for the mouse to move from the light to the dark compartment, was recorded for up to 3 min.

2.7. Histopathological Analysis

On day 35, all mice were deeply anesthetized with 4% isoflurane and perfused transcardially with 0.1 M phosphate‐buffered saline (PBS, pH 7.4). Brains were then removed and fixed in PBS (pH 7.4) containing 4% paraformaldehyde at room temperature for 24 h for hematoxylin and eosin (H&E) and immunohistochemical staining (n ≥ 3). After fixation, the tissues were dehydrated using a graded ethanol series, cleared in alcohol–benzene and xylene, and embedded in paraffin. Paraffin‐embedded tissues were sectioned at 2 μm thickness and stained with H&E or 0.1% cresyl violet (Nissl staining) for microscopic examination. Histopathological changes in the brain were observed using a light microscope (Leica Microsystems Ltd., Wetzlar, Germany). In the hippocampal CA1 region, the number of intact pyramidal neurons, defined as cells with a distinct nucleus and nucleolus, was counted along 1.35‐mm linear segments at 50× magnification. Neuronal density was expressed as the mean number of pyramidal cells per square millimeter (mm²) from three coronal sections per group. For quantitative histopathological analysis, intact pyramidal neurons in the hippocampal CA1 region were counted according to predefined morphological criteria, including a clearly visible nucleus and nucleolus. Three coronal sections per animal were analyzed, and neuronal density was calculated as the mean number of intact pyramidal neurons per square millimeter. Although the original analysis was not performed under fully blinded conditions, identical anatomical regions, section numbers, and morphological criteria were applied across all groups to minimize analytical bias.

2.8. Western Blotting

Proteins (30 μg) were separated using 10–15% SDS polyacrylamide gels and transferred to PVDF membranes. For low molecular weight peptides, 100 μg samples and 16% Tricine SDS gels (Koma Biotech, Seoul, Republic of Korea) were used. The membranes were blocked in TBS‐T containing 5% skim milk for 2 h at room temperature and attached at 4°C overnight with the primary antibodies as follows: β‐amyloid for APP (#51‐2700, 1:1,000), Nrf2 (#PA5‐27882, 1:1,000), Keap1 (#PA5‐99434, 1:2000), IL‐1β (#P420B, 1:1,000), and β‐actin (#PA1‐183, 1:500) from Thermo Fischer Scientific, and Aβ₄₂ (#14974, Aβ(1‐42), D9A3A), Bax (#2772, 1:1000), Bcl‐2 (#3498, 1:1000), brain‐derived neurotrophic factor (BDNF) (#60071, 1:1000), Caspase‐3 (#9662, 1:1000) and β‐actin (#3700, 1:5000) from Cell Signaling Technology (Danvers, MA, USA). The membrane was incubated with the primary antibody mixed with horseradish‐peroxidase‐conjugated goat anti‐rabbit IgG (H + L) secondary antibody (#31460, 1:10,000) for 1 h at room temperature (Oh et al. 2020). The membrane was detected using MicroChemi (version 4.2; DNR Bio‐Imaging Systems Ltd., Neve Yamin, Israel) with a chemiluminescence solution. For densitometric analysis, Western blot band intensities were quantified using ImageJ software. The intensity of each target protein band was normalized to the corresponding β‐actin band from the same sample. The normalized values were then expressed relative to the normal control or scopolamine‐treated group, as indicated in each figure legend. Although the original densitometric analysis was not performed under fully blinded conditions, the same exposure settings, background subtraction procedure, and normalization method were consistently applied to all groups.

2.9. Metabolite Profiling

2.9.1. Mass Spectrometry‐Based Metabolite Profiling Analysis

Organic acids (OAs) and fatty acids (FAs) from mouse serum and brain regions were analyzed using a GCMS‐TQ8040 triple quadrupole system (Shimadzu, Kyoto, Japan) equipped with an Ultra‐2 column (25 m × 0.20 mm, 0.11 μm film thickness; Agilent, USA). Brain samples from the cortex and hippocampus were each processed separately. The cortex was homogenized at a final concentration of 250 mg/mL, whereas the hippocampus was prepared at 100 mg/mL. Both tissues were disrupted on ice using an ultrasonic homogenizer to obtain uniform lysates. The oven program initiated at 100°C for 2 min, increased at 10°C/min to 300°C, and was maintained for 8 min. Serving as the carrier gas, helium was added at a flow rate of 0.5 mL/min, and argon was used for the collision processes, with electron ionization at 70 eV. Derivatization was performed using a methoximation–TBDMS procedure. Serum and tissue extracts were mixed with distilled water supplemented with 3,4‐dimethoxybenzoic acid and pentadecanoic acid (0.1 μg each) as internal standards, after which methoxyamine hydrochloride (1 mg) was added. After adjusting pH to ≥ 12 with 5 M NaOH, samples were incubated at 60°C for 1 h. Acidification to pH ≤ 2 with 10% sulfuric acid was performed before sequential extraction with diethyl ether and ethyl acetate. The organic phases were combined, treated with triethylamine, and concentrated to dryness at 40°C under nitrogen. The residue was taken up in toluene and derivatized using MTBSTFA at 60°C for 1 h for subsequent GC–MS/MS analysis. AAs, nucleosides, kynurenine metabolites, and neurotransmitters in cortex, hippocampus, and serum were quantified using an LCMS‐8050 triple quadrupole mass spectrometer (Shimadzu, Kyoto, Japan) equipped with an Intrada Amino Acid column (50 mm × 3.0 mm, 3 μm). Electrospray ionization was performed with a nebulizing gas flow of 3.0 L/min, heating gas at 10 L/min, an interface temperature of 300°C, and a desolvation line temperature of 250°C. Mobile phase composition depended on the metabolite class: AA separation was performed using 0.1% formic acid in acetonitrile as solvent A and 100 mM ammonium formate in water as solvent B, while nucleoside analysis employed 0.1% formic acid in water as solvent A and 0.1% formic acid in water:acetonitrile (ACN) (4:6) as solvent B. All other LC‐MS/MS metabolite analyses used 0.1% formic acid in water (A) and 0.1% formic acid in ACN (B). For the LC–MS/MS analysis of amino acids, nucleosides, and related metabolites, serum and brain extracts were processed using a protein precipitation procedure optimized for this study. Briefly, 20 μL of each serum sample or brain homogenate was combined with 60 μL of ACN containing a mixture of ISs ¹³C₁‐phenylalanine (25 ng), 3‐deazauridine (2.5 ng), and serotonin‐d₄ (10 ng). The mixture was vortexed for 3 min to ensure complete precipitation of the proteins and subsequently centrifuged. The supernatant was then transferred to a clean vial, filtered, and evaporated to dryness under a gentle stream of nitrogen. The dried residue was then reconstituted in 50 μL of 0.1% formic acid in distilled water, and the clarified solution was placed into autosampler vials for LC–MS/MS analysis.

2.9.2. Star Pattern Recognition Analysis

Star pattern recognition analysis was performed to visualize the relative changes in metabolite abundance among groups. Metabolite concentrations were quantified from the calibration curves generated for each analyte, and the mean values obtained from the experimental groups were normalized to the corresponding mean concentrations of the control group to obtain the relative expression ratios. The normalized data were then imported into Microsoft Excel 2010, and star symbol plots were constructed to depict the comparative metabolic profiles across groups.

2.9.3. Multivariate Statistical Analysis

Partial least squares‐discriminant analysis (PLS‐DA) and pathway enrichment analysis were applied as multivariate approaches using MetaboAnalyst (version 6.0). The optimal number of components was determined based on the cross‐validation results, and the corresponding R² and Q² values were reported. To assess the possibility of model overfitting, permutation testing was performed using 1,000 random permutations of class labels. The performance of the original PLS‐DA model was compared with that of the permuted models, and permutation p‐values were used to evaluate model robustness. Prior to statistical processing, all metabolite data were subjected to log10 transformation followed by autoscaling to standardize the variance across variables.

2.10. Statistical Analysis

All data were expressed as the mean ± SD. Statistical analysis was performed using GraphPad Prism 10 statistical software. Differences among groups were analyzed using one‐way analysis of variance (ANOVA), followed by Dunnett's multiple comparisons test. Statistical significance was set at p < 0.05.

3. Results and Discussion

3.1. Synthesis of Novel Amine‐Substituted Naphthoquinone Derivatives

Consistent with our goal to identify potent Aβ aggregation inhibitors, we systematically modified the 1,4‐naphthoquinone core by introducing diverse amine functionalities at the C2 position. The synthetic route involved a CeCl₃·7H₂O‐catalyzed coupling between commercially available 1,4‐dihydroxynaphthalene and various amines in ethanol. This approach efficiently afforded a library of 2‐amino‐substituted naphthoquinones, as exemplified by the synthesis of compound 888 using a trimetazidine‐derived piperazine partner (35% yield; Scheme 1).

Scheme 1.

Schematic showing the synthesis of compound 888 (TPN).

Using this established protocol, we successfully obtained additional key derivatives, including 587, 591, and 671 (Figure 2) with yields ranging from 38% to 82%. For example, the reaction with tryptamine produced 587 with a high yield of 82%, incorporating a rigid indole‐ethyl motif characteristic of endogenous neuro‐signaling molecules. To explore the effects of the hydrophobic bulk material, 591 was prepared by introducing a longer aliphatic side chain using N,N‐dibutylaminopropylamine (38% yield). Furthermore, the incorporation of an aromatic amine linkage was realized in compound 671 (80% yield) through a reaction with N,N‐dimethyl‐p‐phenylenediamine, allowing the evaluation of the impact of electronic conjugation on biological efficacy. Their ¹H‐NMR, ¹³C‐NMR, and mass spectra are provided in the Supplementary Information (Figures S1–13). These targeted structural variations at the C2 position provided a necessary control library to systematically investigate how altered electronic properties, steric bulk, and flexibility influence the core scaffold's therapeutic potential against Alzheimer's pathology.

Figure 2.

Chemical structures of representative Aβ aggregation inhibitors (587, 591, 671, and 888 (TPN)).

3.2. Biochemistry

3.2.1. Aβ Aggregation Inhibition Studies of Hit Compounds

The synthesized naphthoquinone derivatives were subsequently screened for their ability to inhibit the aggregation of Aβ peptides, the primary pathological hallmark of AD. Screening results revealed that the inhibitory potential was highly dependent on the nature of the amine substituent.

After 54 synthetic compounds of quinone and naphtoquinone derivatives were tested, four hit compounds, 587, 591, 671, and TPN, were selected as Aβ aggregation inhibitors for further analysis in this study. The IC₅₀ values of the four compounds were 2.31 ± 0.037, 1.94 ± 0.093, 1.91 ± 0.055, 0.14 ± 0.004 μM, respectively (Table 1). Compound TPN showed the best inhibitory activity with dose‐dependent inhibition (Figure 3). The inhibitory activity of compound TPN for Aβ aggregation was 11.64 times higher than that of curcumin (IC₅₀ = 1.63 ± 0.14 μM), while the inhibitory activity of 671 and 591 was slightly lower than that of curcumin. In addition, the potency of TPN was higher than a naphthoquinone compound 2 (IC₅₀ = 3.2 μM for Aβ₄₀) (Campora et al. 2021b).

Table 1.

Inhibitory activities of the hit compounds against Aβ aggregation and related enzymes.

Compound	IC50 (μM)
	Aβ aggregation	BACE1	AChE	BChE	MAO‐A	MAO‐B
587	2.31 ± 0.04	> 40	13.44 ± 0.54	> 40	11.73 ± 0.36	0.028 ± 0.003
591	1.94 ± 0.09	> 40	12.77 ± 0.67	> 40	5.53 ± 1.15	10.93 ± 0.78
671	1.91 ± 0.06	> 40	10.45 ± 0.31	7.61 ± 0.31	> 40	0.0070 ± 0.0001
TPN	0.14 ± 0.005	> 40	> 40	13.64 ± 0.41	8.37 ± 0.78	14.38 ± 0.23
Curcumin	1.63 ± 0.14	—	—	—	—	—

—, not determined.

AChE, acetylcholinesterase; BACE1, β‐site APP cleaving enzyme 1; BChE, butyrylcholinesterase; MAO‐A and MAO‐B, monoamine oxidases A and B, respectively.

Figure 3.

IC₅₀ curves of TPN and curcumin at seven different concentrations of each compound.

Notably, the compounds exhibiting the most significant Aβ aggregation inhibitory activity were those coupled with specific amine partners: tryptamine (587), 3‐(dibutylamino)propylamine (591), N,N‐dimethyl‐p‐phenylenediamine (671), and 2,3,4‐trimethoxybenzylpiperazine (TPN). The pronounced activity of these particular derivatives suggests that the presence of certain structural motifs—such as the indole ring in 587 (tryptamine), the tertiary amine in the lipophilic side chain of 591, the aromatic diamine structure of 671, and the substituted piperazine ring of TPN—is crucial for effective binding and interference with Aβ self‐assembly. When comparing these hit compounds, although 587, 591, and 671 maintained reasonable inhibitory profiles similar to the reference compound curcumin (IC₅₀ = 1.63 μM), they lacked the capability to strongly lock into the amyloid peptide over time. As revealed later by our molecular dynamics simulations (Section 3.3.2), compound like 671 completely lost atomic contact with the protein during the initial stages of simulation, mirroring the short‐lived interactions observed for curcumin. Conversely, only compound TPN reestablished prolonged and stable contacts within the regulatory regions of Aβ₄₂, proving that the trimetazidine‐derived piperazine ring provides an irreplaceable structural and thermodynamic advantage over the aliphatic, indole, or single‐aromatic variants found in 587, 591, and 671. This comparative SAR discussion underscores why TPN was uniquely selected as the ultimate lead candidate for further in vivo translational evaluations.

3.2.2. BACE1, AChE, BChE, and MAO Inhibition Studies of the Hit Compounds as MTDLs

The hit compounds were tested for MTDL activities for BACE1, AChE, BChE, MAO‐A, and MAO‐B. Although the four compounds exhibited weak inhibitory activities against BACE1 (IC₅₀ > 40 μM), these showed potent or effective inhibitory activities against MAO‐B (Table 1). Compound TPN showed effective inhibitory activities against MAO‐A (IC₅₀ = 8.37 μM), MAO‐B (IC₅₀ = 14.38 μM), and BChE (IC₅₀ = 13.64 μM), and 671 exhibited strong inhibition against MAO‐B (IC₅₀ = 0.0070 μM) and effective inhibition against BChE (IC₅₀ = 7.61 μM) and AChE (IC₅₀ = 10.45 μM). In addition, compound 591 had effective inhibitory activities against MAO‐A (IC₅₀ = 5.53 μM), MAO‐B (IC₅₀ = 10.93 μM), and AChE (IC₅₀ = 12.77 μM), and 587 possessed potent MAO‐B inhibitory activity (IC₅₀ = 0.028 μM), and effective inhibitory activity against MAO‐A (IC₅₀ = 11.73 μM) and AChE (IC₅₀ = 13.44 μM) (Table 1). These results suggest potent Aβ aggregation inhibitors TPN and 671 are candidates for multitarget‐directed ligands (MTDLs), similar to naphthoquinone 2, which inhibits AChE (IC₅₀ = 9.2 μM) and MAO‐B (0.0077 μM) (Campora et al. 2021b).

3.3. Molecular Docking and Dynamics Simulation

3.3.1. Molecular Docking of Hit Compounds

The docking scores calculated using AutoDock Vina and the interacting residues identified by PLIP for the four leading molecules with curcumin are shown in Table 2. Among the five docked molecules, TPN was predicted to have the highest binding affinity for the peptide. The compound showed a stronger predicted binding affinity (–6.325 kcal/mol) than curcumin (–5.929 kcal/mol). The four identified molecules interacted with the residues in the EFRH (residues 3–7) motif through a hydrogen bond with ASP7, a target for monoclonal antibodies (Gardberg et al. 2009). Compound TPN also interacted with GLN15 and GLU22, the starting and end positions of the potential β‐strand region (15‐21) (Hsu et al. 2018). GLU22 is also known as one of the key residues in Aβ₄₂ fibrilization kinetics (Hsu et al. 2018). The predicted binding poses of TPN, 671, and curcumin are shown in Figure 4.

Table 2.

Docking scores of the four hit molecules and curcumin predicted by AutoDock Vina.

Compound	Binding affinity (kcal/mol)	Interacting residues
587	−6.131	TYR10, GLU11, HIS14, GLN15
591	−5.073	ASP7, TYR10, GLU11, HIS14, GLN15, VAL18
671	−6.036	ASP7, TYR10, GLU11, HIS14
TPN	−6.325	ASP7, TYR10, HIS14, GLN15, GLU22
Curcumin	−5.929	TYR10, HIS14, GLN15

Figure 4.

Binding poses of (A) TPN, (B) 671, and (C) curcumin to the peptide Aβ₄₂. The ligands are represented as the ball and stick representation. The atomic structures of interacting residues identified by PLIP are shown.

3.3.2. Molecular Dynamics Simulation of Hit Compounds

During the MD simulation, we counted the number of atomic contacts with the three ligands and Aβ₄₂ peptide. The distance cutoff to determine all‐atom contact was set to 5.0 Å (Figure 5). Given the intrinsically disordered protein nature of the peptide and thus high flexibility, all three molecules lost contacts with the peptide at the initial stage of MD simulation. The number of contacts became zero at 3.6 ns, 7.6 ns, and 2.3 ns for TPN, 671, and curcumin, respectively. However, the compound TPN reestablished the contacts with the protein during the 22‐31 ns. After the contacts with the peptide were lost for the three ligands, no other contacts were observed by the end of the simulation, 200 ns. This indicates that the molecule remained around the binding residues, potentially impeding aggregation. On the other hand, compounds 671 and curcumin did not make any contact with the protein again after they lost the interactions. The time that each ligand contacts with the peptide correlated with the predicted binding affinity by AutoDock Vina. Compound TPN, showing the longest contact time, had the lowest IC₅₀ value of 0.14 μM, while 671 and curcumin had same order of IC₅₀, 1.91 μM and 1.63 μM, respectively.

Figure 5.

Number of contacts during molecular dynamics (MD) simulation between the peptide Aβ₄₂ and compound TPN (blue), 671 (green), and curcumin (red).

3.4. In Silico Pharmacokinetics of Leading Compounds

In the in silico pharmacokinetics prediction, all four compounds were predicted to have high GI absorption and BBB permeability, suggesting their potential as CNS‐active drug candidates for the CNS. In addition, none of these compounds were predicted to be P‐gp substrates, suggesting that they maintain stable intracellular concentrations as drug candidates. Although all compounds predicted low skin permeability and compounds 587 and 671 predicted inhibitory effects on CYP450 enzymes, compounds 591 and 888 were predicted to be relatively stable, suggesting their potential suitability for oral or intravenous administration (Table 3). These results suggest the potential applicability of these four compounds as drug candidates, especially compounds 591 and 888, which may have the potential for further development as drug candidates.

Table 3.

In silico pharmacokinetics properties of the hit compounds.

Compound	GI absorption	BBB permeant	P‐gp substrate	Inhibitor					Log Kp
				CYP 1A2	CYP 2C19	CYP 2C9	CYP 2D6	CYP 3A4	Skin permeation (cm/s)
587	High	Yes	No	Yes	Yes	Yes	Yes	Yes	−5.58
591	High	Yes	No	No	Yes	No	Yes	Yes	−5.01
671	High	Yes	No	Yes	Yes	Yes	Yes	Yes	−5.98
TPN	High	Yes	No	No	Yes	Yes	Yes	Yes	−5.81

BBB, blood‐brain barrier; GI, gastrointestinal; P‐gp, P‐glycoprotein.

3.5. Cytotoxicity and Neuroprotectivity

The cytotoxicity of TPN and 671 in MDCK and SH‐SY5Y cells was evaluated using the CCK‐8 assay. Compounds TPN and 671 were nontoxic to MDCK cells. However, in SH‐SY5Y cells, compound 671 exhibited a low toxicity with 77.67% viability at 30 μM, whereas TPN was not toxic at any concentrations (Figure 6).

Figure 6.

Cell viability of MDCK and SH‐SY5Y cells after treatment of compounds TPN and 671. C, Control; D, dimethyl sulfoxide (DMSO). Cell viability was measured after 24 h of culture using the CCK‐8 assay (n = 5). The final concentration of DMSO was maintained below 1% in all experiments.

To evaluate the protective effect of the leading compound TPN against Aβ₄₂‐induced toxicity, SH‐SY5Y cells were treated with Aβ₄₂ or TPN for 24 h. Aβ₄₂ treatment significantly reduced cell viability. However, co‐treatment with TPN significantly restored the viability of Aβ₄₂‐treated SH‐SY5Y cells (Figure 7).

Figure 7.

Effect of co‐treatment with Aβ₄₂ and the leading compound TPN on SH‐SY5Y cell viability. The final concentration of Aβ₄₂ and TPN was 10 μM. Cell viability was measured using the CCK‐8 assay after 24 h of culture. The value of control was used to normalize the data. Data are shown as the mean ± standard deviation (SD) (n = 5). ****p < 0.0001.

3.6. Animal Behavioral Tests

Scopolamine, a muscarinic antagonist, is widely used in AD research because it induces oxidative stress, increases Aβ and phosphorylated tau levels (Tang 2019), and impairs memory in both humans and rodents. In contrast, donepezil (DPZ), despite causing several adverse effects such as diarrhea, nausea, vomiting, dizziness, insomnia, and vivid dreams, similar to other AD therapeutics, remains the most widely prescribed drug owing to its applicability across all stages of AD and its relatively long‐term use (Wahl et al. 2017). Therefore, in this study, DPZ was used as a positive control to evaluate the efficacy of TPN in a mouse model of SCO‐induced cognitive and memory impairment.

In the MWM test, which assesses semantic and spatial working memory in rodents, the escape latency to locate the hidden platform progressively decreased in the normal control (NOR) group, whereas the scopolamine‐treated control (SCO) group showed little improvement compared to day 1. Notably, TPN‐treated mice exhibited a progressive reduction in escape latency, similar to the NOR group. On day 5, the escape latency in the NOR group was 15.1 ± 9.7 s, whereas the SCO group showed a significantly prolonged latency of 50.8 ± 8.6 s. In contrast, TPN treatment markedly reduced the latency to 21.3 ± 8.3 s, while the positive control (SCO + DPZ) group exhibited a latency of 46.7 ± 13.9 s, indicating that TPN improved long‐term spatial memory (Figure 8A–C).

Figure 8.

Effects of TPN on spatial memory impairment in scopolamine‐treated mice. (A) escape latency for 5 days; (B) Mice arriving at the platform in a Morris water maze (MWM) test; (C) representative tracks of the test during the probe trial on day 5; (D) step‐through latency in passive avoidance (PA) test; and (E) Alternation Triplet (%) in Y‐maze test. NOR, normal control group; SCO, scopolamine‐treated group; TPN, scopolamine + TPN‐treated group; DPZ, scopolamine + donepezil‐treated positive control group. Data are presented as the mean ± standard error of the mean (SEM) (n = 8 per group). Statistical significance was determined by one‐way ANOVA followed by Tukey's post hoc test, except for escape latency during the MWM training period, which was analyzed by two‐way repeated‐measures ANOVA followed by Tukey's post hoc test. *p < 0.05, **p < 0.01, and ***p < 0.001 versus NOR group; ^# p < 0.05, ^## p < 0.01, and ^### p < 0.001 versus SCO group; ns, not significant.

The passive avoidance test, which relies on the amygdala in the medial temporal lobe and is widely used to evaluate long‐term memory (Arvanitakis et al. 2019), further confirmed the memory‐restorative effects of TPN. In this paradigm, the innate preference of rodents for dark compartments is challenged by mild electrical stimuli to assess their inhibitory control in avoiding an aversive environment. Step‐through latency in the bright compartment was 60.1 ± 7.1 s in the NOR group, but was significantly reduced to 24.5 ± 8.3 s in the SCO group, confirming scopolamine‐induced amnesia. TPN treatment restored memory function, as evidenced by an increased latency of 60.1 ± 16.9 s to the level of NOR group, whereas DPZ showed a latency of 33.8 ± 8.8 s (Figure 8D).

Finally, in the Y‐maze test, a simple behavioral assay for short‐term working memory (Gacar et al. 2011), the alternation behavior of the SCO group was significantly reduced to 42.9 ± 4.3%, compared with 51.3 ± 5.7% in the NOR group. TPN treatment significantly increased alternation behavior to 49.9 ± 2.7%, suggesting enhanced memory performance. Importantly, there were no significant differences in the total number of arm entries among the groups, indicating that the observed improvements were due to memory enhancement rather than changes in locomotor activity (Figure 8E).

3.7. Histopathological Analysis

To evaluate the neuroprotective effects of TPN on morphological changes and neuronal damage in SCO hippocampal damage, brain tissue was collected and subjected to H&E and Nissl staining. H&E staining revealed that pyramidal neurons in the hippocampal CA3 and dentate gyrus (DG) regions in the SCO group exhibited morphological alterations and reduced cell numbers due to damage (Figure 9A,B). In contrast, TPN administration preserved the structural integrity of hippocampal neurons, which remained clearly distinguishable. Consistent with this, Nissl staining demonstrated that neurons in the NOR group displayed intact cellular structures with distinct morphologies and dense, orderly arrangements (Figure 9C,D). Conversely, in the SCO group, pyramidal neurons in the CA1, CA3, and DG regions exhibited features such as nuclear fragmentation, clumping, and pyknosis, as well as ghost cell formation, accompanied by reduced cell density and lower tissue compactness. Notably, in the TPN‐treated group, most pyramidal neurons in the hippocampal region maintained their typical morphology and structural integrity, indicating that TPN effectively protected the hippocampal neurons.

Figure 9.

Effects of TPN on hippocampal neurons. (A) H&E staining of mouse hippocampi. (B) The proportions of damaged cells in the hippocampal DG/CA1/CA3 regions assessed by H&E staining. (C) Nissl staining of mouse hippocampi. Images were taken through a compound light microscope at 40× (Scale bars = 500 μm) and 400× (Scale bars = 50 μm) magnification. (D) The proportions of damaged cells in the hippocampal DG/CA1/CA3 regions assessed by Nissl staining.

3.8. Western Blotting

Western blot analyses were performed using dissected hippocampal and cortical tissues isolated from treated mice after the animal behavioral tests. APP, as a full length Aβ, and Aβ₄₂ were primarily targeted in Western blotting since the drug TPN was selected as a Aβ aggregation inhibitor. In addition, expression levels of Nrf2 and keap1 as oxidative stress factors, IL‐1β as an inflammation factor, and Bax, Bcl2, BDNF, and caspase‐3 as neuron apoptosis factors were analyzed.

In the experiments, Aβ₄₂ was quantified with the band corresponding to its oligomeric form (~27 kDa) after the Tricine‐SDS‐PAGE. The oligomeric band was observed in the analysis when the pure Aβ₄₂ peptide was solely loaded (Figure S14). As results, APP and Aβ₄₂ levels in the SCO group were increased compared to NOR group, but TPN group had decreased levels in hippocampus and cortex; however, those levels were ameliorated to levels similar to that of the positive DPZ group (Figure 10). These results suggested that TPN influenced on precursor protein APP formation and shorter peptide Aβ₄₂. The changes might come from the differences that monomers of APP or Aβ₄₂ are more easily metabolized than their aggregates. These findings are similar to those of menadione, a naphthoquinone derivative (Zhang et al. 2018). However, osmundacetone decreased the Aβ₄₂ level but did not significantly alter the APP levels (Zhao et al. 2024). The TPN group decreased the keap1 level and increased Nrf2 levels in the hippocampus and cortex, compared to the SCO group (Figure 10). Similar, TPN significantly decreased IL‐1β level in hippocampus, compared to the SCO group, and showed comparable levels to positive‐control group (Figure 10). These results suggested that TPN alleviated oxidative stress by upregulating Nrf2 and inflammation by downregulating IL‐1β expression.

Figure 10.

Effect of TPN on Aβ aggregation, oxidative stress, and inflammation in the brain of scopolamine (SCO)‐induced mice. (A) Western blotting of the cortex; (B) quantification of protein expression of the cortex; (C) Western blotting of the hippocampus; (D) and quantification of protein expression of the hippocampus. The values were normalized to β‐actin using ImageJ software. Supernatant of each homogenate was subjected to SDS‐PAGE, and Western blotting analysis was performed using each specific antibody against Aβ (APP), Aβ₄₂, IL‐1β, Keap1, and Nrf2. *Aβ₄₂ was quantified with its oligomer form (~27 kDa). NOR, normal control; SCO, scopolamine‐induced group; TPN, TPN‐treated group; DPZ, donepezil‐treated positive group. Data in the histogram are expressed as means ± SD of three independent experiments (n = 3). *p < 0.05, **p < 0.01 and ***p < 0.001 versus NOR or SCO groups as indicated by bars.

In addition, the expression levels of apoptosis‐related proteins were analyzed to evaluate the neuroprotective effect of TPN against scopolamine‐induced apoptosis (Figure 11A–D). As shown in Figure 10A,C, scopolamine administration significantly increased the expression of pro‐apoptotic proteins Bax and cleaved caspase‐3 and markedly decreased the expression of the anti‐apoptotic protein Bcl‐2 compared with the normal control group. In contrast, TPN treatment effectively attenuated the scopolamine‐induced apoptotic signaling. Specifically, TPN administration restored Bcl‐2 expression and suppressed Bax expression in both the cortex (Figure 10B) and the hippocampus (Figure 10D), indicating a reduction in neuronal apoptosis. In addition to its anti‐apoptotic effects, the potential mechanism by which TPN improves cognitive function and memory was investigated by examining the expression of brain‐derived neurotrophic factor (BDNF). As shown in Figure 10C,D, SCO treatment significantly reduced the BDNF protein levels in the cortex. However, this reduction was significantly attenuated by TPN administration in a dose‐dependent manner, suggesting that TPN attenuated SCO‐induced impairment of neurotrophic support.

Figure 11.

Effect of TPN on apoptosis and associated neuronal proliferation in the brain of scopolamine‐induced mice. (A) Western blotting of the cortex; (B) quantification of protein expression of the cortex; (C) Western blotting of the hippocampus; (D) and quantification of protein expression of the hippocampus. The values were normalized to β‐actin using ImageJ software. Supernatant of each homogenate was subjected to SDS‐PAGE, and Western blotting analysis was performed using each specific antibody against Bax, Bcl‐2, BDNF, and Caspase‐3. Data in the histogram are expressed as the means ± SD of three independent experiments (n = 3 per group). *p < 0.05, **p < 0.01 and ***p < 0.001 versus NOR or SCO groups.

3.9. Metabolomics Study

A total of 98 metabolites were identified in the serum, 91 in the cortex, and 89 in the hippocampus. Among these, 11 metabolites in the serum, 19 in the cortex, and 23 in the hippocampus were significantly altered in the NOR and SCO groups (p < 0.05). For clarity, only representative metabolites illustrating the predominant SCO‐related trends are described in the main text, whereas the full list of significantly altered metabolites is provided in Supplementary Tables S1–S3.

Star symbol plots based on fold changes (SCO/NOR) revealed clear SCO‐induced metabolic deviations across the serum, cortex, and hippocampus. In the SCO/NOR comparison, the serum showed increased succinic acid and DOPA levels and decreased 3‐hydroxykynurenine levels (Figure S15). The cortex exhibited elevated isoleucine and reduced GABA levels (Figure S16). In the hippocampus, increased lactate and reduced citric acid levels indicated a glycolytic‐TCA shift (Figure S17). These representative alterations were consistent with the metabolites identified in the univariate analysis, confirming the reliable metabolic reflection of SCO across tissues.

The drug‐treated groups exhibited a partial but distinct attenuation of SCO‐associated deviations. In the serum, TPN partially normalized the changes in 3‐hydroxykynurenine and DOPA, suggesting a relatively selective modulation of amino acid and tryptophan‐related metabolism, whereas DPZ showed a broader corrective tendency that also involved energy metabolism linked metabolites such as succinic acid (Figure S15). Prior metabolomics studies have similarly indicated that SCO perturbs serum metabolic profiles and that donepezil‐containing regimens can modulate systemic metabolic signatures in scopolamine‐induced cognitive impairment models (Kim et al. 2024). In the cortex, both TPN and DPZ moderated SCO‐associated changes in isoleucine and GABA, which are metabolites commonly annotated in SCO‐related brain metabolomics profiling (Figure S16) (Yoon et al. 2024). In the hippocampus, SCO‐induced disruptions in central energy metabolism (e.g., lactate and citric acid) were minimally corrected by TPN but were more effectively attenuated by DPZ, in line with metabolomic profiling in SCO‐treated mice where energy‐related metabolites (including lactate/citrate/succinate) contributed considerably to group discrimination (Figure S17) (Yoon et al. 2024). This pattern is consistent with the evidence that DPZ can attenuate mitochondrial dysfunction (Ye et al. 2015). Moreover, SCO models have been linked to altered hippocampal mitochondrial integrity and tryptophan‐kynurenine pathway enzyme dynamics, providing a biological context for the kynurenine‐related signals observed in this study (Herbet et al. 2025).

The PLS‐DA score plots showed a partial but consistent separation between the NOR and SCO groups across the serum, cortex, and hippocampus (Figure 12). The models were validated using a five‐fold cross‐validation. All models demonstrated a good fit (R² ranging from 0.655 to 0.861), and predictive capabilities varied across tissues: serum (five components, R² = 0.861, Q² = 0.458), cortex (three components, R² = 0.823, Q² = 0.540), and hippocampus (five components, R² = 0.865, Q² = 0.668). To assess potential overfitting of the PLS‐DA models, permutation testing was performed using 1,000 permutations. The original PLS‐DA models showed statistically significant performance compared with the permuted models in serum, cortex, and hippocampus (serum, p = 0.001; cortex, p = 0.013; hippocampus, p = 0.024), supporting the reliability of the observed group separation.

Figure 12.

Partial least squares‐discriminant analysis (PLS‐DA) score plots of metabolite profiles for the NOR, SCO, TPN, and DPZ groups in (A) serum, (B) cortex, and (C) hippocampus.

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis (p < 0.05, impact score > 0.1) highlighted sample‐specific pathway perturbations. Pathways were prioritized using a composite score (–log10(p‐value) × impact score) to integrate statistical significance and pathway topology (Xia and Wishart 2011). Overall, the serum showed relatively focused changes with SCO versus TPN (tryptophan/histidine/pyrimidine metabolism), whereas SCO versus DPZ additionally emphasized arginine biosynthesis and broader amino acid/energy‐related pathways (e.g., alanine–aspartate–glutamate, arginine–proline, and glyoxylate/dicarboxylate) (Figure 13A, Table 4). In the cortex, TPN mainly affected alanine–aspartate–glutamate metabolism with TCA‐cycle involvement, whereas DPZ produced a wider shift, spanning lipid (linoleic acid metabolism), aromatic amino acid pathways, and energy pathways (TCA cycle and pyruvate metabolism) (Figure 13B, Table 4). In the hippocampus, SCO versus TPN was associated with broad amino acid/redox/energy metabolism (e.g., arginine biosynthesis, arginine–proline, glutathione, TCA cycle, and glyoxylate/dicarboxylate), whereas SCO versus DPZ was limited to a smaller subset (alanine–aspartate–glutamate, glycine–serine–threonine, tyrosine metabolism, and one‐carbon pool by folate) (Figure 13C, Table 4).

Figure 13.

Metabolic pathway analysis for pairwise group comparisons in (A) serum, (B) cortex, and (C) hippocampi. Each bubble plot displays pathway enrichment (–log10 p‐value) on the y‐axis and pathway impact values derived from pathway topology analysis on the x‐axis. Bubble size represents pathway impact, and bubble color indicates the statistical significance of each pathway.

Table 4.

Sample‐specific Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways ranked by integrated significance and impact scores after TPN and DPZ treatment in scopolamine‐exposed mice.

	Pathway name	p‐value	−log (p‐value)	Impact score	Composite scorea
Serum
SCO vs TPN	Pathway name	p‐value	−log(p‐value)	Impact score	Composite scorea
	Tryptophan metabolism	0.019	1.724	0.53	0.91
	Histidine metabolism	0.036	1.445	0.22	0.32
	Pyrimidine metabolism	0.009	2.064	0.11	0.22
SCO vs DPZ	Arginine biosynthesis	< 0.001	3.489	0.51	1.76
	Alanine, aspartate and glutamate metabolism	0.009	2.070	0.54	1.12
	Histidine metabolism	< 0.001	3.779	0.22	0.84
	Tryptophan metabolism	0.044	1.355	0.53	0.72
	Arginine and proline metabolism	0.024	1.625	0.40	0.65
	Pyrimidine metabolism	< 0.001	4.351	0.11	0.46
	Glyoxylate and dicarboxylate metabolism	0.018	1.747	0.17	0.29
	Butanoate metabolism	0.045	1.345	0.14	0.19
	Lysine degradation	0.043	1.368	0.11	0.15
Cortex
SCO vs TPN
	Alanine, aspartate and glutamate metabolism	0.008	2.084	0.76	1.59
	Arginine biosynthesis	0.037	1.428	0.51	0.72
	Citrate cycle (TCA cycle)	0.041	1.390	0.51	0.72
	Cysteine and methionine metabolism	0.025	1.602	0.24	0.39
SCO vs DPZ	Alanine, aspartate and glutamate metabolism	< 0.001	6.620	0.76	5.05
	Linoleic acid metabolism	0.009	2.058	1.00	2.06
	Phenylalanine, tyrosine and tryptophan biosynthesis	0.018	1.737	1.00	1.74
	Tyrosine metabolism	< 0.001	4.172	0.40	1.69
	Citrate cycle (TCA cycle)	0.001	2.991	0.51	1.54
	Glycine, serine and threonine metabolism	0.010	1.981	0.51	1.02
	Cysteine and methionine metabolism	0.001	3.203	0.24	0.78
	Pyruvate metabolism	< 0.001	3.511	0.22	0.78
	Phenylalanine metabolism	0.018	1.737	0.36	0.62
	Glyoxylate and dicarboxylate metabolism	0.007	2.177	0.25	0.54
	One carbon pool by folate	0.002	2.752	0.11	0.29
Hippocampus
SCO vs TPN
	Alanine, aspartate and glutamate metabolism	0.001	3.189	0.76	2.43
	Arginine biosynthesis	< 0.001	4.524	0.51	2.29
	Arginine and proline metabolism	< 0.001	5.106	0.40	2.04
	Glycine, serine and threonine metabolism	< 0.001	3.321	0.51	1.70
	Glutathione metabolism	< 0.001	3.486	0.37	1.31
	Citrate cycle (TCA cycle)	0.010	2.001	0.51	1.03
	Glyoxylate and dicarboxylate metabolism	< 0.001	3.450	0.25	0.86
	Cysteine and methionine metabolism	0.001	3.280	0.24	0.80
	Tyrosine metabolism	0.042	1.380	0.40	0.56
	Histidine metabolism	0.005	2.298	0.22	0.51
	One carbon pool by folate	< 0.001	3.371	0.11	0.36
	Pyruvate metabolism	0.040	1.394	0.22	0.31
	Pyrimidine metabolism	0.006	2.245	0.11	0.24
	Lysine degradation	0.028	1.553	0.11	0.17
SCO vs DPZ	Alanine, aspartate and glutamate metabolism	0.042	1.374	0.76	1.05
	Glycine, serine and threonine metabolism	0.013	1.892	0.51	0.97
	Tyrosine metabolism	0.013	1.901	0.40	0.77
	Cysteine and methionine metabolism	0.028	1.548	0.24	0.38
	One carbon pool by folate	0.014	1.868	0.11	0.20

^a –log10(p‐value) × impact score.

Taken together, these analyses demonstrate that SCO induces characteristic region‐dependent metabolic disturbances across the serum, cortex, and hippocampus, encompassing neurotransmitter‐related metabolites and central energy metabolism. TPN and DPZ partially counteracted these SCO‐associated deviations; however, their metabolic footprints were distinct. In the serum and cortex, TPN exhibited relatively selective modulation, particularly affecting amino acid‐ and tryptophan‐related metabolism, whereas DPZ showed broader normalization, extending to energy‐ and lipid‐associated pathways. In contrast to the serum and cortex, the hippocampus showed the opposite pattern: pathway perturbations were more extensive in the SCO versus TPN comparison (including arginine‐related, glutathione, and TCA/energy‐associated pathways), whereas the SCO versus DPZ comparison was characterized by a more restricted subset of pathway changes. These treatment‐specific patterns were consistently supported by fold‐change profiles, PLS‐DA clustering, and composite‐ranked pathway enrichment, indicating that TPN and DPZ mitigated SCO‐induced metabolic dysregulation through overlapping yet differentiable metabolic pathways.

Overall, these findings should be interpreted within a hierarchical mechanistic framework. The ThT assay and molecular docking/MD simulation provide direct evidence supporting the anti‐Aβ aggregation activity of TPN. The behavioral improvement, reduced APP/Aβ42‐related protein expression, attenuation of inflammatory, oxidative stress‐related, and apoptotic markers, preservation of hippocampal neurons, and changes in BDNF expression provide supportive in vivo evidence of neuroprotection under scopolamine‐induced impairment conditions. In contrast, the metabolomic changes, particularly those involving amino acid and kynurenine metabolism, should be considered correlative evidence reflecting treatment‐associated metabolic responses. Thus, the present results support TPN as a multifunctional neuroprotective lead compound, while avoiding the interpretation that all datasets directly prove an anti‐Aβ mechanism.

Although the present findings demonstrate that TPN improved cognitive performance and exerted neuroprotective effects in the scopolamine‐induced cognitive impairment model, this pharmacological model has important limitations in terms of translational relevance to Alzheimer's disease pathology. The scopolamine model is useful for evaluating cholinergic dysfunction and memory impairment; however, it does not fully reproduce the progressive Aβ plaque deposition, tau pathology, and chronic neurodegeneration observed in Alzheimer's disease. Therefore, the present in vivo results should be interpreted as evidence of cognitive improvement and neuroprotection under scopolamine‐induced impairment conditions, rather than as direct proof of disease‐modifying anti‐amyloid activity in Alzheimer's disease. Further studies using Aβ‐driven transgenic AD models, such as APP/PS1 or 5xFAD mice, are required to confirm whether TPN can suppress amyloid pathology and exert disease‐modifying effects in vivo. A further limitation is the difference in administration routes between TPN and donepezil. TPN was administered intraperitoneally, whereas donepezil was administered orally and was used as a positive reference control rather than as a pharmacokinetic‐equivalent comparator. Therefore, direct efficacy comparison between TPN and donepezil should be interpreted with caution, because route‐dependent differences in absorption, first‐pass metabolism, systemic exposure, brain exposure, and peak concentration may have influenced the results. Future studies using the same administration route, together with pharmacokinetic and brain distribution analyses of TPN, are required.

4. Conclusion

In this study, a synthesized naphthoquinone derivative, we referred to as compound TPN, was found to be a potent Aβ aggregation inhibitor and improved cognitive functions in SCO‐induced mice, and significantly changed the metabolite profiling of brain tissues. Compound TPN showed the highest Aβ aggregation inhibitory activity (IC₅₀ = 0.14 ± 0.005 μM) with better potency than the reference curcumin (IC₅₀ = 1.63 ± 0.14 μM). Compound TPN showed effective inhibitory activities against MAO‐A, MAO‐B, and BChE, and 671 exhibited potent inhibition against MAO‐B (IC₅₀ = 0.0070 ± 0.0001 μM), and effective inhibitory activities against BChE and AChE, suggesting TPN and 671 can be candidates for MTDLs. Both compounds were predicted to be BBB‐permeable and nontoxic to MDCK and SH‐SY5Y cells, except 671 exhibited a low toxicity to SH‐SY5Y cells. The final selected leading compound TPN was predicted to have the lowest binding affinity and maintained contact with the protein during the MD simulation, contributing to inhibition of the Aβ aggregation. TPN significantly improved cognitive functions in behavioral tests (MWM, PAT, and Y‐maze) on SCO‐induced mice. In Western blotting and histopathological analysis, TPN‐treatment significantly decreased APP as Aβ, Aβ₄₂, IL‐1β level, Keap1, and anti‐apoptosis factors and appeared to preserve most pyramidal neurons in the CA1 region, which largely retained their typical morphology and structural integrity. In metabolite profiling, TPN generally modulated a relatively narrower set of pathways than DPZ, showing mainly amino acid‐ and kynurenine‐related modulation, whereas DPZ induced broader adjustments spanning amino acid, mitochondrial/energy, and lipid‐related pathways particularly in serum and cortex; in contrast, hippocampal pathway perturbations were more extensive in the SCO versus TPN comparison than in SCO versus DPZ comparison. Although the present findings demonstrate that TPN improved cognitive performance and exerted neuroprotective effects in the scopolamine‐induced cognitive impairment model, this pharmacological model has important limitations in terms of translational relevance to Alzheimer's disease pathology. The scopolamine model is useful for evaluating cholinergic dysfunction and memory impairment; however, it does not fully reproduce the progressive Aβ plaque deposition, tau pathology, and chronic neurodegeneration observed in Alzheimer's disease. Therefore, the present in vivo results should be interpreted as evidence of cognitive improvement and neuroprotection under scopolamine‐induced impairment conditions, rather than as direct proof of disease‐modifying anti‐amyloid activity in Alzheimer's disease. Further studies using Aβ‐driven transgenic AD models, such as APP/PS1 or 5xFAD mice, are required to confirm whether TPN can suppress amyloid pathology and exert disease‐modifying effects in vivo.

Collectively, BBB‐permeable TPN showed cognitive‐protective and neuroprotective effects under scopolamine‐induced impairment conditions, accompanied by attenuation of inflammatory responses, preservation of neuronal structures, and treatment‐associated metabolic changes. These findings support TPN as a multifunctional neuroprotective lead compound for further investigation in neurodegenerative disease‐related model.

Author Contribution

Jong Min Oh: Formal analysis, investigation, writing – original draft. Won Kyeong Jeong: Formal analysis, investigation. Hyun Ju Son: Formal analysis, investigation. Seo Young Kim: Formal analysis, investigation, writing – original draft. Tae Woo Oh: Conceptualization, writing – review and editing, supervision, funding acquisition. Moongi Ji: Formal analysis, investigation, writing – original draft. Ji Woo Nam: Formal analysis, investigation, writing – original draft. Minyeong Baek: Formal analysis, investigation, writing – original draft. Jeong‐Ho Park: Conceptualization, writing – review and editing, supervision. Woong‐Hee Shin: Writing – review and editing, supervision. Hee Jung Kim: Formal analysis, investigation. Byeongchan Choi: Formal analysis, investigation, writing – original draft. Man‐Jeong Paik: Conceptualization, writing – review and editing, supervision. Hoon Kim: Conceptualization, writing – review and editing, supervision, project administration, funding acquisition.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Data will be made available on request.