Multi-Target Effects of Novel Synthetic Coumarin Derivatives Protecting Aβ-GFP SH-SY5Y Cells against Aβ Toxicity

Alzheimer’s disease (AD) is a common neurodegenerative disease presenting with progressive memory and cognitive impairments. One of the pathogenic mechanisms of AD is attributed to the aggregation of misfolded amyloid β (Aβ), which induces neurotoxicity by reducing the expression of brain-derived neurotrophic factor (BDNF) and its high-affinity receptor tropomyosin-related kinase B (TRKB) and increasing oxidative stress, caspase-1, and acetylcholinesterase (AChE) activities. Here, we have found the potential of two novel synthetic coumarin derivatives, ZN014 and ZN015, for the inhibition of Aβ and neuroprotection in SH-SY5Y neuroblastoma cell models for AD. In SH-SY5Y cells expressing the GFP-tagged Aβ-folding reporter, both ZN compounds reduced Aβ aggregation, oxidative stress, activities of caspase-1 and AChE, as well as increased neurite outgrowth. By activating TRKB-mediated extracellular signal-regulated kinase (ERK) and AKT serine/threonine kinase 1 (AKT) signaling, these two ZN compounds also upregulated the cAMP-response-element binding protein (CREB) and its downstream BDNF and anti-apoptotic B-cell lymphoma 2 (BCL2). Knockdown of TRKB attenuated the neuroprotective effects of ZN014 and ZN015. A parallel artificial membrane permeability assay showed that ZN014 and ZN015 could be characterized as blood–brain barrier permeable. Our results suggest ZN014 and ZN015 as novel therapeutic candidates for AD and demonstrate that ZN014 and ZN015 reduce Aβ neurotoxicity via pleiotropic mechanisms.


Introduction
Alzheimer's disease (AD), the most prevalent type of neurodegenerative dementia, is characterized by progressive memory and cognitive impairments [1]. Extracellular accumulation of misfolded amyloid β (Aβ) in the brain (amyloid plaques) contributes to neuronal apoptosis, eventually leading to the shrinkage of the cortex and hippocampus. Aβ is produced from the cleavage of amyloid peptide precursor protein (APP) by βand γ-secretases [2]. The Aβ tends to form oligomers and fibrils, which cause neuronal death by increasing oxidative stress, neuroinflammation, excitotoxicity, and apoptosis [3]. Among these mechanisms, Aβ-induced oxidative stress modifies proteins to perturb their biological function and impairs key biochemical and metabolic pathways in which these proteins normally play a role [4]. In addition, selective loss of acetylcholine-containing neurons in the brain contributes substantially to the cognitive decline in AD [5], and acetylcholinesterase (AChE) inhibitors modulating acetylcholine hydrolysis can increase the level and action duration of acetylcholine [6]. Accumulation of Aβ has also been proposed to be an activator to induce sequential pathological events such as the downregulation of the brain-derived neurotrophic factor (BDNF) signaling pathway [7,8].
BDNF, a member of the neurotrophic factor family, regulates the survival and differentiation of neurons by binding to its high-affinity receptor tropomyosin-related kinase B (TRKB) [9]. The binding of BDNF to TRKB induces the dimerization and autophosphorylation of TRKB [10] to activate the downstream extracellular signal-regulated kinase (ERK) and AKT serine/threonine kinase 1 (AKT). The phosphorylation of the cAMP-responseelement binding protein (CREB) by ERK and AKT [11,12] further upregulates expressions of BDNF [13] and anti-apoptotic B-cell lymphoma 2 (BCL2) [14]. BCL2 binds to apoptosis regulator BCL2-associated X (BAX) to inhibit BAX-mediated mitochondrial outer membrane permeabilization, thereby inhibiting apoptosis [15,16]. The accumulation of oligomeric Aβ downregulates BDNF expression [17] and impairs the retrograde axonal transport of TRKB [18]. Intracerebral injection of BDNF in animal models of AD reduces Aβ-induced neurotoxicity and synaptic loss and improves memory impairments [19]. Therefore, the potentiation of the BDNF signaling pathway by TRKB agonists would be a strategy in treating AD.
Coumarins belong to a family of oxygen-containing heterocycles with a scaffold of 1,2benzopyrone. These compounds exhibit diverse pharmacological effects such as reducing inflammation and oxidative stress and have been widely used as complementary and alternative medicines in treating neurodegenerative diseases [20]. It has been reported that derivatives of coumarin could prevent misfolded Aβ aggregation [21]. In AD cell and mouse models, synthetic coumarin-chalcone hybrid LM-031 demonstrates neuroprotective potential by regulating CREB and anti-oxidative pathways [22,23]. Coumarin derivative imperatorin also activates BDNF and CREB signaling to improve learning and memory deficits in prenatally stressed rats [24]. In addition, osthole lessens cognitive impairment in estrogen-deficiency mice by rescuing the reduction of BDNF and TRKB, as well as phosphorylation of CREB, in the hippocampus [25]. Here, we report the potential of two newly synthetic coumarins, ZN014 and ZN015, to reduce Aβ aggregations and oxidative stress as well as to enhance the TRKB signaling pathway in SH-SY5Y neuroblastoma cell models for AD.

Thioflavin T Aggregation Assay
To form Aβ aggregation, Aβ 42 peptide (10 µM; Kelowna Int'l Scientific Inc., New Taipei City, Taiwan) was incubated with tested compounds (5-20 µM) in 150 mM NaCl and 20 mM Tris-HCl (pH8.0) at 37 • C for 48 h. Thioflavin T (10 µM; Sigma-Aldrich) was added to the Aβ mixture and incubated at room temperature for 5 min. The fluorescence intensity was recorded at 420 nm excitation and 485 nm emission by an FLx800 microplate reader (Bio-Tek, Winooski, VT, USA). Half maximal effective concentration (EC 50 ) was estimated by a method of interpolation.
To examine ROS, CellROX Deep Red reagent (5 µM; Molecular Probes, Eugene, OR, USA) was added to the cells and incubated at 37 • C for 30 min. ROS in cells was measured with excitation 631/28 nm and emission 692/40 nm wavelengths.

Neurite Outgrowth Analysis
Aβ-GFP SH-SY5Y cells (6 × 10 4 cells) were seeded on a 24-well plate with retinoic acid (10 µM) on day 1. Tested coumarins (5 µM) and doxycycline (5 µg/mL) were added on day 2, as described. On day 8, the cells were washed with phosphate-buffered saline (PBS) twice and fixed in 4% paraformaldehyde at 4 • C for 15 min. Cells were permeabilized with Triton X-100 (0.1%) for 10 min, blocked with bovine serum albumin (3%) for 20 min, and stained with anti-neuronal TUBB3 antibody (1:1000; Covance, Princeton, NJ, USA) at 4 • C overnight. The next day, cells were washed with PBS twice and stained by a secondary donkey antirabbit Alexa Fluor ® 555 antibody (1:1000; Thermo Fisher Scientific) and 4 ,6-diamidino-2phenylindole (DAPI; 0.1 µg/mL; Sigma-Aldrich) at room temperature for 1 h. Neuronal images from at least 60 individual fields (150-250 neurons per field) per experiment were captured at excitation 531/40 nm and emission 593/40 nm wavelengths using an ImageXpress micro confocal high-content imaging system (Molecular Devices). Neurite total length (µm) and numbers of process (the number of primary neurites originated from the cell body of a neuron) and branch (the number of secondary neurites extended from primary neurites) were analyzed using a MetaXpress neurite outgrowth application module (Molecular Devices). For each sample, around 6000 cells were analyzed in each of three independent experiments.
To measure AChE activity, cells were suspended in cold PBS and lysed by sonication. After centrifugation, the supernatants were collected. AChE activity in 10 µg protein extracts was measured using an AChE activity assay kit (Sigma-Aldrich). The absorbance of the colorimetric product was measured at 412 nm wavelength (Multiskan TM GO spectrophotometer; Thermo Fisher Scientific).  Taiwan) antibody. The immune complexes were detected using goat anti-mouse or goat anti-rabbit IgG antibody conjugated with horseradish peroxidase (1:5000; GeneTex, Irvine, CA, USA) and chemiluminescent substrate (Millipore).

RNA Interference
Lentiviral short hairpin RNA (shRNA) targeting TRKB (TRCN0000002243, TRCN 0000002245, and TRCN0000002246) and a negative-scrambled control (TRC2.Void) were obtained from the National RNAi Core Facility, IMB/GRC, Academia Sinica (Taipei, Taiwan). As described, cells were plated on 6-or 24-well plates, with retinoic acid added on day 1. Cells were infected with lentivirus (multiplicity of infection, 3 for each shRNA), with polybrene (8 µg/mL; Sigma-Aldrich) on the next day. Cells were pretreated with tested compounds (5 µM) for 8 h after changing medium, followed by doxycycline on day 3. Cells were collected for further analysis on day 9. The hairpin sequences of targeting shRNA were below:

Parallel Artificial Membrane Permeability Assay (PAMPA)
PAMPA was used to predict the penetration of the tested compounds across the BBB. Briefly, the donor well (Millipore) was filled with 300 µL of the tested compound (1 µM) and QC compounds (carbamazepine, theophylline, or lucifer yellow, 100 µg/mL; Sigma-Aldrich). The filter PVDF membrane (pore size 0.45 µm; Millipore) was coated with 4 µL of porcine polar brain lipid (20 mg/mL; Avanti Polar Lipids, Alabaster, AL, USA) in dodecane and the acceptor well filled with 200 µL of 5% DMSO in PBS. The filter plate was carefully placed on the donor plate to form a sandwich plate at room temperature for 18 h. After the permeation time, the filter and donor plates were separated. The concentration of the tested compound in the donor and acceptor wells was measured by an AB Sciex QTrap 5500 mass spectrometer (Applied Biosystems) linked to a 1200 HPLC system (Agilent Technologies, Palo Alto, CA, USA). The concentrations of the QC compounds were determined by a Tecan Infinite M200 Pro microplate reader (Switzerland). The effective permeability coefficient (P e ) was calculated as described [30]. Each compound was tested in triplicate.

Statistical Analysis
All experiments were in triplicate. Data are presented as mean ± standard deviation. Differences between groups were evaluated by a two-tailed Student's t-test or one-way analysis of variance with a post hoc Tukey test where appropriate. The level of statistical significance was expressed as a p-value less than 0.05.

Inhibition Aβ Aggregation and Oxidative Stress by Coumarin Derivatives
SH-SY5Y cells expressing the Aβ-GFP folding reporter [28] were used to evaluate the Aβ aggregation-inhibitory effects of the tested compounds. In these cells, the misfolding and aggregation of Aβ affect the folding of GFP and reduce its fluorescent signal, while the inhibition of Aβ aggregation improves GFP folding and, thereby, increases the fluorescent signal [36]. The Aβ-GFP SH-SY5Y cells were differentiated for 7 days [37], with the induction of Aβ-GFP expression by doxycycline for 6 days (Figure 2A). Under the condition of plating cells and the addition of retinoic acid on day 1 to induce neuronal differentiation, no increased cell density was observed. In addition, neither doxycycline addition nor Aβ induction obviously affected cell viability. As the treatment of curcumin at 10 µM led to appreciable cell death (viability below 80%), 0.2-5 µM concentrations of the compounds, typically in 5-fold dilutions, were selected. After normalization, with cell number counted, treatment with curcumin (111-128%), ZN014 (111-119%), or ZN015 (113-125%) at 1-5 µM significantly increased the GFP fluorescence intensity compared with untreated cells (100%) (p = 0.028-0.001) ( Figure 2B). No significant change of cell viability was detected (111-96%; p > 0.05). Treatment with curcumin, ZN014, or ZN015 at 5 µM did not significantly affect the relative Aβ-GFP/HPRT1 RNA level (29.2-30.0 vs. 28.6 folds of induction) ( Figure 2C).

Discussion
Up to the present, effective therapy to slow the progression of neurodegeneration in AD remains an unmet need. Analogs of coumarins showing pharmacological activities have been described [49]. Coumarin and its derivatives demonstrate their potential in treating AD through several mechanisms such as inhibiting AChE [50] and β-secretase [51], preventing misfolded Aβ aggregation [21], upregulating CREB and anti-oxidative pathways [22,23], and promoting BDNF-TRKB and CREB signaling [24,25]. Here, we found the potential of new coumarin derivatives ZN014 and ZN015 for AD treatment by reducing Aβ aggregation, ROS, caspase-1, and AChE as well as promoting neurite outgrowth (Figures 2 and 3) and TRKB signaling (Figure 4). Knockdown of TRKB expression counteracted the neuroprotective effects of these compounds against Aβ toxicity ( Figure 5), demonstrating the neuroprotective mechanism of ZN014 and ZN015 is mediated by enhancing TRKB signaling. It is noted that the knockdown of TRKB did not increase the activity of caspase-1 and AChE. These may be explained by the fact that caspase-1 and AChE activity are elevated mainly by other mechanisms such as increased oxidative stress and inflammation and not by decreased TRKB in the SH-SY5Y cells expressing Aβ-GFP. Our study results are supported by a previous study that has also shown AChE activity is not affected by deficient TRKB [52]. Moreover, the partial neurite outgrowth rescue effects of ZN014 and ZA015 in cells with knockdown of TRKB also indicate the contribution of other signaling pathways to the neuroprotection of these compounds.
Oxidative stress has been identified as an important factor contributing to the neurodegeneration of AD [53]. Compounds with anti-oxidative potential may directly serve as chemical chaperones to suppress protein aggregates, quench free oxygen radicals, or enhance anti-oxidative signaling to influence cellular ROS [54,55]. In our study, only ZN015 displayed chemical chaperone activity for Aβ aggregation ( Figure 1B), and both ZN014 and ZN015 showed no 1,1-diphenyl-2-picrylhydrazyl radical scavenging activity against ROS (data not shown). As coumarin and its derivatives demonstrate the potential to activate NRF2 anti-oxidative signaling in different cells and animal models [22,56], the antioxidative effect of ZN014 and ZN015 in our cell model ( Figure 2E) may also be upregulated by anti-oxidative signaling.
The production of ROS by Aβ aggregation upregulates caspase-1 activity and induces neuroinflammation [57]. Caspase-1 is involved in the cleavage and activation of interleukins 1β, 18, and 33 [58]. In addition, caspase-1 induces caspase-6 activation, leading to axonal degeneration [39], and axonopathy is recognized as an early event of patients with AD [59]. Axonal degeneration, with swellings of haphazardly arranged vesicles, mitochondria, multilamellar bodies, and vacuoles, and impaired axonal transport could be observed to precede the development of amyloid plaques in the Tg-swAPP Prp mouse model for AD [59]. Activation of caspase-1 also induces pyroptosis with the secretion of TNF-α and IL-6 [60]. Inhibition of caspase-1 reverses memory impairment and decreases Aβ accumulations and neuroinflammation in the brains of the caspase-1 null J20 mouse model of AD [40]. The coumarin derivative nodakenin has been reported to inhibit the production of cytokines via the suppression of caspase-1 activation in anaphylactic mice [61]. In our study, both ZN014 and ZN015 counteracted the Aβ-induced increase in caspase-1 activity ( Figure 3A).
AChE, an enzyme breaking down acetylcholine into acetate and choline, also accelerates the formation of Aβ fibrils [62]. ACE inhibitors may improve AD neurodegeneration by increasing the level and action duration of acetylcholine [63] as well as reducing the formation of Aβ aggregation [64]. AChE-inhibitory activities of coumarin derivatives have been reported [50]. Resembling coumarin and LM-031 [23], ZN015 exhibited inhibitory activity on both AChE ( Figure 3B) and Aβ aggregation ( Figure 1B).
Upon BDNF binding, TRKB dimerizes and phosphorylates to initiate intracellular ERK and AKT signaling, leading to CREB phosphorylation for the survival of neurons [65]. Upon the phosphorylation of serine at position 133 (S133), phosphor-CREB translocates to the nucleus and binds to a cAMP-response-element (CRE) [66], thereby inducing the expression of CRE-mediated transcription of genes such as neurotrophic BDNF and BCL2 for neuroprotection [67]. BCL2 prevents BAX redistribution to the mitochondria, where it forms oligomers, resulting in the efflux of cytochrome c and the induction of the apoptotic cascade [68]. In human neurons, Aβ downregulates BCL2 and increases the level of BAX [69]. In Aβ-GFP SH-SY5Y cells, induction of Aβ-GFP expression downregulated BCL2 and upregulated BAX, and ZN014 and ZN015 counteracted changes in gene expression for these CREB-responsive genes (Figure 4). Of note, ZN014 and ZN015 also upregulated the expression of BDNF (Figure 4), forming positive feedback in the BDNF-TRKB-CREB signaling pathway.
Finally, it is well noted that pre-conditioning cellular protection through NRF2 antioxidative signaling has the hormesis feature [70]. Hormesis is an adaptive biological response to drugs or treatment, which indicates that a greater magnitude of therapeutic effect was seen at the middle dose range and a less protective effect, with stronger cell toxicity, was seen at the higher doses of a compound (a specific pattern of biphasic doseresponse of a compound) [71]. The hormesis of anti-oxidative gene networks in redox reactions is also important for dose optimization in treating neurodegenerative diseases [72]. Further study will be needed to explore the interplay between antioxidant signaling and other signals by coumarin derivatives.

Conclusions
In conclusion, we found the neuroprotective potential of two new coumarin derivatives, ZN014 and ZN015, against Aβ neurotoxicity via the inhibition of oxidative stress, caspase-1, and AChE activities and the activation of TRKB signaling in the Aβ-GFP SH-SY5Y cell model ( Figure 6). As AD has complex neurodegenerative pathogenesis, the pleiotropic mechanism of ZN014 and ZN015 make these compounds promising for drug development. However, the SH-SY5Y cell model only emphasizes the degeneration of neurons, while the pathogenesis of AD also involves glial cells such as astrocytes and microglia. The interactions between neurons and glial cells are also not addressed in this cell model. Although ZN014 and ZN015 rescued the neurite outgrowth deficit after Aβ induction, we did not show if those compounds had a neurotrophic effect on neurite outgrowth without Aβ induction. Given that in clinical practice, we will not treat healthy individuals with drugs, we consider that the effects of the compounds on neurite outgrowth in cells without Aβ induction may not be crucial and the experiment could have been skipped in this study. Furthermore, our findings are limited in human cell models. Future validation in AD animal models will be conducted. The binding of ZN014 and ZN015 to TRKB will also be measured using surface plasmon resonance to consolidate their properties as TRKB agonists.

Data Availability Statement:
The data generated during the study are available from the corresponding author upon request.