Alzheimer’s disease (AD) is the most common neurodegenerative disorder accompanied with progressive memory loss and cognitive impairment. The neuropathological diagnosis of AD relies on the presence of amyloid plaques and neurofibrillary tangles (NFTs) (Hardy, 2006). Glycogen synthase kinase-3β (GSK-3β) is an evolutionarily conserved serine/threonine kinase that phosphorylates cellular substrates and thereby regulates diverse cellular functions, including cellular signaling (Grimes and Jope, 2001) and apoptosis (Li et al., 2000). It has been shown that GSK-3β is a crucial contributor to Tau phosphorylation (Lee et al., 2003) and also has been proposed to contribute to amyloid production (Su et al., 2004). In AD brains, GSK-3β is co-localized with neurofibrillary changes, with increased levels in post-synaptosomal supernatant as compared to the controls (Pei et al., 1997). Overexpression of GSK-3β in adult mouse brain results in hyperphosphorylation of Tau in hippocampal neurons and neurodegeneration (Lucas et al., 2001). Inhibition of GSK-3β by BTA-EG₄, an amyloid-binding drug which reduces Aβ-induced toxicity in vitro (Inbar et al., 2006), reduces Tau phosphorylation in brain slice cultures from 3×Tg-AD mice (Croft et al., 2017). Therefore, identification of effective GSK-3β inhibitors modulating aberrant Tau phosphorylation serves as an attractive therapeutic strategy for AD.

There are two isoforms of GSK-3, GSK-3α (51 kDa) and GSK-3β (47 kDa), encoded by separate genes, which share extensive similarities in their catalytic domains, but differ in their N- and C-terminal regions (Woodgett, 1990). Unlike other protein kinases, GSK-3 is constitutively active and is inactivated by upstream regulators in response to stimuli. Several signaling pathways regulate GSK-3 activity (Hur and Zhou, 2010). Activation of the phosphatidylinositol 3-kinase (PI3K)-AKT serine/threonine kinase 1 (AKT) pathway results in inactivation of GSK-3 through phosphorylation at N-terminal Ser9 (GSK-3β) or Ser21 (GSK-3α) (Cross et al., 1995). Inhibitory phosphorylation of GSK-3β at Ser9 can also be achieved through protein kinase C (PKC)-α-induced extracellular signal-regulated kinase (ERK) and mitogen-activated protein kinase 14 (P38) pathways to attenuate Tau phosphorylation at Ser396 (Nemoto et al., 2010). In addition to GSK-3β, Tau phosphorylation can also be mediated by P38 (Lauretti and Praticò, 2015) and mitogen-activated protein kinase 8 (JNK) (Ploia et al., 2011).

Over the past decade, various classes of small molecule GSK-3β inhibitors with diverse mechanisms have been reported for the treatment of AD. Lithium, a natural, inorganic and water-soluble GSK-3β inhibitor, reduces Tau phosphorylation and insoluble Tau levels in transgenic mice overexpressing human P301L mutant Tau (Noble et al., 2005). Although upregulation of brain derived neurotrophic factor might be part of a neuroprotective effect of lithium in AD patients (Leyhe et al., 2009), no treatment effect on GSK-3 activity or cerebrospinal fluid-based biomarker concentrations was observed in lithium trial in AD (Hampel et al., 2009). Another GSK-3β inhibitor, tideglusib, showed safety and efficacy demonstrated in a small pilot study (del Ser et al., 2013). However, a short term (26 weeks) phase II trial of tideglusib produced no clinical benefit in AD (Lovestone et al., 2015). Furthermore, for arylindolemaleimides SB-216763 and SB-415286 that inhibit GSK-3 in vitro, a previous study demonstrated that these compounds promote survival of primary neurons following the survival factor withdrawal or inhibition of PI3K activity (Cross et al., 2001). Although both SB inhibitors strongly facilitate regrowth of cerebellar granule neurons and improve axon regeneration in injured cortical neurons (Seira et al., 2010), SB-216763 inhibited neurite growth in cerebellar and dorsal root ganglion neurons (Alabed et al., 2010). With these collective current results of GSK-3 inhibition, nevertheless, GSK-3 could be considered as a target for modulating production and integration of new neurons in hippocampus as a treatment for neurodegenerative diseases (Eldar-Finkelman and Martinez, 2011). Therefore, identification of new lead GSK-3β inhibitors for treating AD continues to spur considerable interest (Shi et al., 2020).

In this study, virtual screening using the docking computation combined with a compound structure similarity search was first utilized to screen compounds in a database in order to identify novel GSK-3β inhibitors. Among the top-ranked compounds, seven compounds were selected and underwent in vitro GSK-3β inhibition and cell-based assays to examine their inhibition strengths and their effects on Tau aggregation. In addition, neuroprotection effect on primary hippocampal neuronal culture was also investigated.

Virtual screening

The GOLD docking program (Verdonk et al., 2003) was used to screen 1.5 million compounds collected from the Enamine database (http://www.enamine.net/). Details of computational protocol have been described in our previous work (Lin et al., 2016). Briefly, the docking protocol for virtual screening was selected among several tested protocols based on the performance of the associated benchmark tests, including selection of protein crystal structures and hydrogen-bond (H-bond) constraints which enhanced screening results for docking ligands against kinases (Perola, 2006). This protocol optimization process yielded the final protocol for virtual screening. The GSK-3β structure in the GSK-3β-ligand crystal structure (PDB code 2OW3) (Zhang et al., 2007) was used for docking. The protocol of protein H-bond constraints was used, and the C=O of Asp133 and N-H of Val135 in the hinge-region of the ATP-binding site were selected as H-bond formation sites such that compounds forming these finger print H-bonds will be biased and favored. The ChemScore scoring function was used to estimate interaction strength between ligands and GSK-3β. A total of 16 compounds from virtual screening were suggested to undergo experimental tests. The computed binding strengths of these compounds are listed in Supplementary Table 1. Among these suggested top-ranked compounds, seven were selected (Fig. 1) due to budget consideration and availability, and obtained from Enamine (Kiev, Ukraine) for enzyme, cell, and mouse hippocampal primary neuron culture tests. The remaining 9 compounds are shown in Supplementary Fig. 1. The selected 7 compounds (Enamine’s ID) were: VB-030 (Z1157726443), 031 (Z1157726458), 032 (Z56849949), 035 (Z773887392), 036 (Z46527339), 037 (Z46595412) and 041 (Z29017365). According to IUPAC nomenclature, the structures of quinoline compounds VB-030 and VB-037 are 2-(pyridin-4-yl)-4-(p-tolyl)quinoline and (E)-4-(3-(2-(5-nitroquinolin-2-yl)vinyl)quinolin-2-yl)morpholine, respectively.

Figure 1. Structures and formulas of the seven selected VB-030, -031, -032, -035, -036, -037 and -041 obtained from virtual screening of the Enamine database.

Bioavailability and BBB permeation prediction

Molecular weight, H-bond donor, H-bond acceptor, octanol-water partition coefficient, and polar surface area of VB compounds were calculated using Internet software ChemDraw (http://www.perkinelmer.com/tw/category/chemdraw/). In addition, BBB permeation scores were computed using an online BBB prediction server (https://www.cbligand.org/BBB/).

GSK-3β kinase assay

The ability of the seven selected VB compounds to inhibit GSK-3β kinase was evaluated, wherein a known GSK-3β inhibitor SB-216763 (Sigma-Aldrich, St Louis, MO, USA) served as a positive control. GSK-3β kinase activity was measured in the presence of test compounds using GSK-3β Kinase Enzyme System (Promega, Madison, WI, USA). Reactions were performed at 30°C for 30 minutes in the 25 µL mixture containing 25 µM ATP, 0.2 mg/mL GSK-3β substrate, 1 ng GSK-3β, and 0.018 µM SB-216763 (44) or VB compounds. Kinase activity data were measured as relative light units (RLU) directly correlated with the amount of ADP produced by using SpectraMax L microplate reader (Molecular Devices, Sunnyvale, CA, USA).

Tet-on ∆K280 Tau_RD-DsRed SH-SY5Y cells and fluorescence intensity analysis

Human SH-SY5Y cells expressing DsRed tagged pro-aggregated mutant (∆K280) of C-terminal repeat domain of Tau (Tau_RD, Gln²⁴⁴-Glu³⁷² of the longest Tau⁴⁴¹ isoform) (Lin et al., 2016) were maintained in medium containing 5 µg/mL blasticidin and 100 µg/mL hygromycin (Invitrogen, Carlsbad, CA, USA). On day 1, cells were seeded into 96-well (2×10⁴/well) plates with retinoic acid (10 µM; Sigma-Aldrich) added to induce neuronal differentiation. On day 2, the cells were pre-treated with test VB compounds (2-10 µM) for 8 h, followed by turning on the hybrid human cytomegalovirus (CMV)/TetO₂ promoter-driving ∆K280 Tau_RD-DsRed expression with doxycycline (2 μg/mL; Sigma-Aldrich). The cells were kept in the medium containing retinoic acid, doxycycline and test VB compounds for 6 days. In addition, congo red and SB-415286 (2-10 µM; Sigma-Aldrich) were included for comparison. On day 8, cells were stained with Hoechst 33342 (0.1 µg/mL; Sigma-Aldrich), cell images were automatically recorded at 531/40 nm excitation and 593/40 nm emission wavelengths (ImageXpress Micro Confocal High-Content Imaging System, Molecular Devices) and analyzed (MetaXpress High-Content Image Acquisition and Analysis Software, Molecular Devices). EC₅₀ (half maximal effective concentration) was estimated by a method of interpolation.

Cell proliferation assay

To evaluate compound cytotoxicity, 2×10⁴ ∆K280 Tau_RD-DsRed SH-SY5Y cells were plated on 96-well dishes with retinoic acid (10 µM) on day 1. On day 2, the cells were treated with the test compounds (1-100 μM) for 8 h, followed by inducing ∆K280 Tau_RD-DsRed expression with doxycycline (2 μg/mL). On day 8, 10 μL of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) (5 mg/mL) was added to the cells at 37°C for 3 h. 100 μL of lysis buffer (10% Triton X-100, 0.1 N HCl, 18% isopropanol) was then added to dishes and the absorbance of the insoluble purple formazan product at OD 570 nm was read by a FLx800 fluorescence microplate spectrophotometer (Bio-Tek, Winooski, VT, USA). IC₅₀ (half maximal inhibitory concentration) was estimated by a method of interpolation.

Real-time quantitative reverse transcription PCR analysis

∆K280 Tau_RD-DsRed SH-SY5Y cells were seeded in a 6-well plate (5×10⁵/well), differentiated with retinoic acid, and treated with congo red, SB-415286, VB-030 or VB-037 (10 µM) and doxycycline as described. On day 8, cells were collected and total RNA was extracted using Trizol reagent (Invitrogen). The RNA was reverse-transcribed using high-capacity cDNA reverse transcriptase (Thermo Fisher Scientific, Waltham, MA, USA). Real-time quantitative PCR experiments were performed using 100 ng cDNA and customized Assays-by-Design probe for DsRed and HPRT1 (4326321E) using StepOnePlus Real-time PCR system (Applied Biosystems, Foster City, CA, USA). Fold change was calculated using the formula 2^∆Ct, ∆C_T=C_T (HPRT1)–C_T (DsRed), in which C_T indicates cycle threshold.

Western blot analysis

As described, ∆K280 Tau_RD-DsRed SH-SY5Y cells were seeded in a 6-well plate and treated with retinoic acid, test compounds (10 µM) and doxycycline. On day 8, cells were collected and total proteins were prepared using lysis buffer containing 10 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM EDTA, pH 8.0, 0.1% SDS, 1% sodium deoxycholate, 1% NP-40, as well as protease (Sigma-Aldrich) and phosphatase (Abcam, Cambridge, MA, USA) inhibitor cocktails. Proteins (25 μg) were separated on 10% SDS-polyacrylamide gel electrophoresis and transferred onto polyvinylidene difluoride membranes (Sigma-Aldrich) by reverse electrophoresis. After blocking, the membrane was probed with DsRed (1:500; Santa Cruz Biotechnology, Dallas, TX, USA), GSK-3α and GSK-3β (1:500; Santa Cruz Biotechnology), p-GSK-3α (S21) (1:1000; Cell Signaling Technology, Beverly, MA, USA), p-GSK-3β (S9) (1:1000; Cell Signaling Technology), ERK and p-ERK (T202/Y204) (1:1000; Cell Signaling Technology), AKT and p-AKT (S473) (1:2000; Cell Signaling Technology), P38 and p-P38 (T180/Y182) (1:1000; Cell Signaling Technology), JNK and p-JNK (T183/Y185) (1:2000; Cell Signaling Technology), Tau (1:500; Dako, Glostrup, Denmark), p-Tau (S202) (1:2000; AnaSpec Inc., Fremont, CA, USA), p-Tau (T231 and S404) (1:2000; Sigma-Aldrich), p-Tau (S396) (1:1000; Sigma-Aldrich), and GAPDH (1:2000; MDBio Inc., Taipei, Taiwan) at 4°C overnight. After extensive washing, the immune complexes were detected by horseradish peroxidase-conjugated goat anti-mouse or goat anti-rabbit IgG antibody (1:5000; GeneTex, Irvine, CA, USA) and chemiluminescence substrate (Millipore, Billerica, MA, USA).

Compound treatment on primary hippocampal neuronal culture

All experiments were performed according to the guidelines of the Institutional Animal Care and Use Committee of National Taiwan Normal University, Taipei, Taiwan. Female pregnant mice (C57BL/6J) were purchased from the National Breeding Centre for Laboratory Animals (Taipei, Taiwan). Mouse hippocampi were isolated from embryonic brains on days 16°C18 and trypsinized (0.05%) for 15 min in 37°C and cells were cultured in neurobasal plating media as described). Briefly, the hippocampal cells were plated into poly-L-lysine (100 μg/mL)-pretreated flasks (48-well plates; 3×10⁴ cells per culture well). The cultures were incubated in a tissue culture incubator at 37°C in 5% CO₂. On days in vitro (DIV) 1, 4 and 7, half of the culture media was replaced with fresh media without horse serum. On DIV 4 and 7, cytosine arabinoside (2 μM) was added to the culture to reduce the glial cell populations, which the hippocampal neuronal population was enriched. On DIV 9, the cells were treated with 10 nM wortmannin and GF109203X to induce Tau hyperphosphorylation to mimic the AD pathological condition (Liu et al., 2003; Xu et al., 2005). VB-030 or VB-037 (1 µM) was added to the cells sequentially. Cells were harvested 12 h later for immunocytochemical staining.

Immunocytochemical staining of hippocampal primary culture

The neuronal survival and morphology of the hippocampal primary culture were evaluated by immunocytochemical staining with antibodies against NeuN (neuron-specific RNA binding nuclear protein) and Map2 (microtubule-associated protein 2, for neurite morphology). Cells were fixed and stained with primary antibodies (1:1000; NeuN and Map2; Millipore) and then fluorescence tagged secondary antibodies and 4’,6-diamidino-2-phenylindole (DAPI) for nuclei. Mature neuron, neurite length and branching were analyzed by Metamorph image analysis software (ImageXpress Micro, Molecular Devices) using images from evenly distributed 16 arenas in each well (totally 64 arenas was set) from independent three experiments. Each neurite length longer than 2 folds of neuronal soma diameter was included.

Statistical analysis

For each set of values, data were expressed as the mean ± standard deviation (SD). Three independent tests in two or three biological replicates were performed in each experiment and differences between groups were evaluated by Student’s t-test or one-way analysis of variance (ANOVA) with post-hoc Tukey test, where appropriate. Statistical significance was set at p<0.05.

Selected top-ranked compounds from virtual screening and prediction of bioavailability and BBB permeation

A large number of structurally diverse potent GSK-3β inhibitors have been reported (Xu et al., 2019). In pursuit of finding potent GSK-3β inhibitors for the treatment of AD, we combined docking calculations and similarity searches to perform structure-based virtual screening. Supplementary Fig. 2 displays GSK-3β structure used for docking and the finger print H-bond interactions between the hinge-region of the ATP-binding site and compounds. Structures and formulas of the selected seven top-ranked VB compounds (030, 031, 032, 035, 036, 037 and 041) are shown in Fig. 1. Based on molecular weight (MW), hydrogen bond donors (HBD), hydrogen bond acceptors (HBA), and calculated octanol/water partition coefficient (cLogP), most VB compounds meet Lipinski’s rule of 5 guidelines for predicting oral bioavailability (MW≤450, HBD ≤5, HBA≤10, cLogP≤5) (Lipinski et al., 2001) (Table 1). With a polar surface area (PSA) less than 90 Ų, compounds except VB-041 were predicted to diffuse across the blood-brain barrier (BBB) (Hitchcock and Pennington, 2006), as also suggested by online BBB predictor (BBB permeation score greater than the threshold 0.02) (Liu et al., 2014) (Table 1).

Table 1 Prediction of bioavailability and BBB permeation

Compound	VB-030	VB-031	VB-032	VB-035	VB-036	VB-037	VB-041
HBD	0	2	3	2	1	0	2
HBA	2	5	5	4	3	7	8
cLogP	5.1	4.6	5.0	5.0	6.1	4.5	3.4
PSA (Å2)	25	71	61	44	46	89	112
BBB score	0.341	0.095	0.193	0.237	0.181	0.096	0.060

MW, molecular weight; HBD, hydrogen bond donor; HBA, hydrogen bond acceptor; cLogP, calculated octanol-water partition coefficient; PSA, polar surface area; BBB, blood-brain barrier.

Test of GSK-3β inhibition

In vitro GSK-3β inhibition assay was used to examine the inhibition strengths of these VB compounds. When the concentrations of these VB compounds were set at 0.018 μM, the IC₅₀ of SB-216763 at 25 µM ATP (Lin et al., 2016), the detected residual activities of GSK-3β were shown in Table 2. Under the reaction conditions of this study, the VB compound concentration of 0.018 µM did not form colloidal aggregates, and thus non-specific enzyme-aggregate inhibition could be ruled out. The lower residual GSK-3β activity indicates stronger inhibitory ability of the compound. Among these VB compounds, VB-030 was the most potent in inhibiting GSK-3β (36.1 ± 1.1% residual GSK-3β activity at 0.018 μM), whereas VB-037 was the least potent in inhibiting GSK-3β (90.9 ± 6.6% residual GSK-3β activity at 0.018 μM).

Table 2 Inhibition potency against GSK-3β kinase

Compound	VB-030	VB-031	VB-032	VB-035	VB-036	VB-037	VB-041
Residual GSK-3β activity at 0.018 μM (%)	36.1 ± 1.1	63.1 ± 1.2	82.0 ± 8.8	79.9 ± 3.8	75.2 ± 4.8	90.9 ± 6.6	79.2 ± 6.2

SB-216763 was used as a positive inhibitor control for GSK-3β activity assay which gave an IC₅₀ of 0.018 μM. The data were obtained from three independent experiments (n=3).

Test of VB compounds reducing tau aggregation in ∆K280 Tau_RD-DsRed-expressing SH-SY5Y cells

Tet-On ∆K280 Tau_RD-DsRed SH-SY5Y cells with DsRed fluorescence reflecting Tau aggregation status (Lin et al., 2016) were used to examine the ability of VB compounds to reduce ∆K280 Tau_RD aggregation and promote Tau-DsRed solubility. Congo red and SB-415286 known to reduce Tau aggregates in HEK293 cells (Lin et al., 2016) were included for comparison. The poorly folded ΔK280 Tau_RD formed aggregates to adversely affect the folding of fused DsRed and thus decrease DsRed fluorescence. The retinoic acid-differentiated SH-SY5Y cells were pretreated with different concentrations of test compounds (2-10 µM) for 8 h, followed by inducing ∆K280 Tau_RD-DsRed expression with doxycycline for 6 days (Fig. 2A). As shown in Fig. 2B, treatment with congo red at 5-10 μM (109-118%, p=0.010-0.001), SB-415286 at 10 μM (110%, p=0.008), VB-030 at 10 μM (109%, p=0.010) or VB-037 at 5-10 μM (113-130%, p=0.006-0.002) increased the DsRed fluorescence intensity significantly compared with untreated cells (100%). As a measure of Tau aggregation inhibition, congo red, SB-415286, VB-030 and VB-037 had EC₅₀ values of 32.9, 47.5, 54.9 and 15.5 μM, respectively (Supplementary Fig. 3A). Greater than 77% of cell viability was observed following treatment with congo red, SB-415286, VB-030 or VB-037 at 10 μM concentration. Based on the cell number analyzed, the IC₅₀ cytotoxicity values of congo red, SB-415286, VB-030 and VB-037 in SH-SY5Y cells expressing ΔK280 Tau_RD-DsRed for 6 days were all greater than 10 μM. Representative DsRed fluorescent images of ΔK280 Tau_RD-expressing cells untreated or treated with congo red, SB-415286, VB-030 or VB-037 (10 μM) are shown in Fig. 2C. When cell viability assay was performed following treatment with 2-50 μM congo red, SB-415286, VB-030 and VB-037, IC₅₀ values were 54.8, 53.2, 44.8 and 66.5 µM, respectively (Fig. 2D, Supplementary Fig. 3B). Addition of 10 μM congo red, SB-415286, VB-030 or VB-037 did not alter the relative ΔK280 Tau_RD-DsRed/HPRT1 RNA level as determined by real-time quantitative reverse transcription PCR (24.3-25.2 vs. 24.2 folds of induction; p>0.05) (Fig. 2E). Significantly increased soluble ΔK280 Tau_RD-DsRed protein was also observed in cells treated with congo red, SB-415286, VB-030 or VB-037 (117-129%, p=0.021-<0.001) compared with untreated cells (100%) (Fig. 2F).

Figure 2. VB compounds screen using the Tet-On ΔK280 TauRD-DsRed SH-SY5Y cell model for AD. (A) Experimental flowchart. On day 1, cells were plated with retinoic acid (RA, 10 μM) added to the culture medium. On day 2, congo red, SB-415286 or VB compounds (2–10 μM) was added to the cells for 8 h, followed by inducing ΔK280 TauRD-DsRed expression with doxycycline (Dox, 2 μg/mL) for 6 days. On day 8, DsRed fluorescence, TauRD-DsRed RNA and soluble TauRD-DsRed protein were measured. (B) Assessment of DsRed fluorescence with test compounds treatment (n=3). The relative DsRed fluorescence of untreated cells (Untr.) was normalized as 100%. p values: comparisons between with and without compound addition (*p<0.05, **p<0.01; two-tailed Student’s t test). Shown below were cell number analyzed in each treatment. (C) Fluorescent images of ΔK280 TauRD-DsRed cells with or without congo red, SB-415286, VB-030 or VB-037 (10 μM) treatment. Nuclei were counterstained with Hoechst 33342 (blue). (D) Cytotoxicity of test compounds examined using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. Cells were treated with test compound (2–50 μM) for cell viability measurement (n=3). To normalize, the relative viability of untreated cells was set at 100%. (E) TauRD-DsRed RNA and (F) soluble TauRD-DsRed protein in ΔK280 TauRD-DsRed-expressing cells untreated or treated with congo red, SB-415286, VB-030 or VB-037 at 10 μM (n=3). p values: comparisons between induced (Dox+) vs. uninduced (Dox-) cells (^###p<0.001), or compound-treated vs. untreated (both are under Dox+) cells (*p<0.05, **p<0.01, ***p<0.001; one-way ANOVA with a post hoc Tukey test).

Tests of VB-030 and VB-037 on inhibiting Tau phosphorylation

We then examined the effect of SB-415286, VB-030 and VB-037 on the levels of p-GSK-3α at Ser21 (to inactivate GSK-3α) and p-GSK-3β at Ser9 (to inactivate GSK-3β) (Fig. 3A) in ∆K280 Tau_RD-DsRed-expressing SH-SY5Y cells by Western blotting. No significant difference was detected in the level of total GSK-3α (92-99% versus 100%, p>0.05) or GSK-3β (107-110% versus 100%, p>0.05) protein between test compound-treated and untreated control. As a positive control, a significant increase in the level of p-GSK-3α (Ser21) (150% versus 105%, p=0.002) and p-GSK-3β (Ser9) (133% versus 105%, p=0.023) was observed in SB-415286-treated cells compared with untreated control. However, neither VB-030 nor VB-037 increased expression of GSK-3α Ser21 (103-112%) or GSK-3β Ser9 (102-106%) (p>0.05).

Figure 3. Effects of VB-030 and VB-037 on GSK-3α, GSK-3β, ERK, AKT, P38, JNK and Tau phosphorylation in ΔK280 TauRD-DsRed SH-SY5Y cells. Western blot analysis of (A) GSK-3α (Ser21), GSK-3β (Ser9), (B) ERK (Thr202/Tyr204), AKT (Ser473), P38 (Thr180/Tyr182), JNK (Thr183/Tyr185), and (C) Tau (Ser202, Thr231, Ser396 and Ser404). GAPDH was used as a loading control (n=3). To normalize, the relative protein of uninduced cells was set at 100%. p values: comparisons between with and without doxycycline addition, or between with and without compound treatment (*p<0.05, **p<0.01 and ***p<0.001) (one-way ANOVA with a post hoc Tukey test).

The effects of SB-415286, VB-030 and VB-037 on the levels of kinases which phosphorylate Tau including ERK, AKT, P38 and JNK were then examined (Fig. 3B). Treatment of SB-415286 increased the level of p-ERK (Thr202/Tyr204) (135% versus 103%, p=0.013) and p-AKT (Ser473) (130% versus 96%, p=0.012), while treatment of VB-037 decreased the level of p-P38 (Thr180/Tyr182) (47% versus 77%, p=0.047). Again, no significant difference was detected in the levels of total ERK (91–101%), AKT (97–103%), P38 (88–97%) and JNK (91–101%) between test compound-treated and untreated control (p>0.05).

We then examined whether SB-415286, VB-030 or VB-037 treatment could inhibit endogenous Tau phosphorylation at Ser202, Thr231, Ser396 and Ser404 in ∆K280 Tau_RD-DsRed-expressing SH-SY5Y cells. As shown in Fig. 3C, the steady-state levels of total Tau showed no significant difference between test compound-treated and untreated cells (100-106% versus 100%, p>0.05). However, the levels of p-Tau in SB-415286, VB-030 or VB-037 compound-treated cells were significantly reduced to 60-29% at Ser202 (p=0.011-<0.001), to 73-27% at Thr231 (p=0.004-<0.001), to 76–57% at Ser396 (p=0.038-0.001), and to 55-34% at Ser404 (p=0.017-<0.001), respectively, compared with untreated cells (100%). Collectively, these results suggest that resembling SB-415286, the two VB compounds inhibit Tau phosphorylation in ∆K280 Tau_RD-DsRed SH-SY5Y cells.

Tests of VB compounds on mouse hippocampal primary neuron culture

To evaluate whether compounds VB-030 and VB-037 are with neuroprotective activity, we further applied 10 μM compounds to mouse hippocampal primary culture under Tau toxicity, given that both compounds at 10 μM reduced Tau protein aggregation in SH-SY5Y cells. It has been reported that co-treatment of wortmannin (WT) (inhibitor of PI3K) and GF109203X (GFX) (inhibitor of PKC) increased GSK-3 activity and decreased cell viability (Xu et al., 2005). Therefore, we tested VB compounds on mouse hippocampal primary culture treated with WT and GFX to intensely activated GSK-3 and enhanced Tau hyperphosphorylation. The neuronal survival and neurite morphology were analyzed by immunocytochemical staining. In vehicle control groups, the neurite length ranged 250-300 µm and the number of branches ranged 15-20; while in WT/GFX group, the neurite length ranged 180-200 µm and the number of branches ranged 8-10. The representative staining images are shown in Fig. 4A. The WT and GFX treatment significantly reduced the neuronal survival (p=0.045), neurite length (p=0.006) and branching (p=0.003) (Fig. 4B). VB-037 has significant protective effect for neuronal survival (p=0.009). Both VB-030 and VB-037 significantly increased the neurite length (p<0.001) and branching (p<0.001). Although VB-030 is not effective as VB-037 in neuronal survival, however, its activity in promoting neurite branching (p=0.022) is significantly better than VB-037 (Fig. 4B).

Figure 4. Effect of VB-030 and VB-037 on mouse hippocampal primary culture under Tau toxicity. (A) The representative immunocytochemical image of cultures. Red, green and blue were staining of NeuN, Map2 and DAPI, respectively. (B) The quantitative results of neuronal survival and neurite morphology from immunocytochemical images (n=3). For normalization, the levels of vehicle-treated cells (control) were set as 100%. p values between WT+GFX-treated versus vehicle-treated (^#), VB-030 or VB-037-treated versus untreated (*), or VB-030-treated versus VB-037-treated (&) were compared. ^# and &: p<0.05, ^## and **: p<0.01, ***: p<0.001 (one-way ANOVA with a post hoc Tukey test).

Effective treatments to slow AD neurodegeneration are still unavailable. GSK-3β, a key player in AD pathophysiology, influences all the major hallmarks of the disease including: amyloid-β production, Tau phosphorylation, memory formation and storage, neuronal growth, and synaptic plasticity (Lauretti et al., 2020). The expanding role of GSK-3β inhibitor suggests GSK-3β, a valuable therapeutic target for AD. In the present study, virtual screening was first performed on compounds from an Enamine’s compound database using structure-based docking computation. This yielded 7 potential compounds which exerted inhibition in the GSK-3β enzyme assay. We then used human SH-SY5Y cells expressing pro-aggregator Tau as a platform to identify two quinoline compounds VB-030 and VB-037 as Tau aggregate reducers. Through inhibiting GSK-3β kinase activity or p-P38 (Thr180/Tyr182), both compounds reduced endogenous Tau phosphorylation at Ser202, Thr231, Ser396 and Ser404 in neuronally differentiated SH-SY5Y cells expressing ΔK280 Tau_RD. Furthermore, both compounds displayed neuroprotective activity on mouse hippocampal primary culture under Tau cytotoxicity (Fig. 4).

Similar to the previously described mouse N2a ∆K280 Tau_RD cell model (Khlistunova et al., 2006), our human SH-SY5Y cell model was readily used to develop hits that prevent Tau aggregation in cells. In addition to reducing ∆K280 Tau_RD-DsRed aggregation, VB-030 and VB-037 also reduced Aβ aggregation in both thioflavin T biochemical assay and Aβ-green fluorescent protein (GFP)-expressing HEK293 AD cell model (Chiu et al., 2019). Whether the possible chemical chaperone activity of VB-030 and VB-037 contributes to the folding and/or stability of pro-aggregant Tau remains to be determined.

Tau phosphorylation is regulated by a balance between Tau kinase and phosphatase activities. In the case where the high number of Tau phosphorylation sites is involved in AD, GSK-3β is a promising target involved with 70% pathological Tau phosphorylation sites (Martin et al., 2013). GSK-3β phosphorylates predominantly Ser/Thr-Pro motifs in Tau protein including Ser202, Thr231, Ser396 and Ser400 (Li and Paudel, 2006). Studies of in vitro phosphorylation assay and NMR spectroscopy reveal that Tau residues Ser-396, Ser-400 and Ser-404 can be phosphorylated by GSK3β alone, and Ser-404 is essential in this process, as its mutation to Ala prevents all activity of GSK3β (Leroy et al., 2010). Phosphorylation of Thr231 by GSK-3β plays an important role in Tau’s hyperphosphorylation and functional regulation (Lin et al., 2007). Although mammals have distinct genes encoding GSK-3 α and β isoforms which share amino acid sequence identity of 97% within their kinase domains (Woodgett, 1990), gene knockout experiments in mice revealed that Thr231/Thr235 and Ser396 on Tau proteins are phosphorylated predominantly by GSK-3β isoform (Soutar et al., 2010). At 0.018 μM concentration, residual GSK-3β activities were 36.1% for VB-030 and 90.9% for VB-037 (Table 2), not reflecting the level of inactive pGSK-3β (Ser9) in cells treated with 10 μM VB-030 (102%) or VB-037 (106%) (Fig. 3A). In addition, stronger inhibition of Tau phosphorylation at Ser202 (52% versus 60%), Thr231 (61% versus 73%), Ser396 (57% versus 76%) and Ser404 (42% versus 55%) was seen with VB-037 compared to VB-030 (Fig. 3C), although not significant. VB-037 has been reported to reduce the phosphor/total ratio of P38 in lipopolysaccharide/interferon-γ-inflamed Aβ-GFP-expressing SH-SY5Y AD cell model (Chiu et al., 2019). The lack of negative correlation between the observed Tau Ser202, Thr231, Ser396 and Ser404 phosphorylation and GSK-3β Ser9 phosphorylation in VB-030/037-treated cells indicates that P38 kinase regulating Tau phosphorylation at these sites may also be affected by VB-037 in ∆K280 Tau_RD-DsRed-expressing SH-SY5Y cells.

Through coordinating multiple proliferation and differentiation signals, GSK-3 is a master regulator of neural progenitor homeostasis (Kim et al., 2009). In addition, there are many studies about the association of GSK-3β and β-catenin in AD pathogenesis. Active GSK-3β has been found in brains staged for AD neurofibrillary changes (Pei et al., 1999), and β-catenin levels are reduced in AD patients (Zhang et al., 1998). Overexpression of GSK-3β in mice causes impairment in spatial learning (Hernandez et al., 2002). Over-activated GSK-3β can downregulate the efficiency of the PS1/N-cadherin/β-catenin complex causing deficiencies in synaptic and neuronal viability that leads to AD pathology (Uemura et al., 2007). Therefore, GSK-3β inhibitors stabilize β-catenin, which presents at synaptic terminals associates with the cytoplasmic domain of cadherin and regulate cell adhesion and influence synaptic size and strength (Murase et al., 2002; Maguschak and Ressler, 2008).

Increasing evidences have suggested that GSK-3β activation-mediated hyperphosphorylation of Tau can induce neuronal apoptosis (Gao et al., 2014; Jiang et al., 2016). Inhibition of GSK-3β by lithium treatment reduced Tau phosphorylation and improved cognitive deficits in animal models of AD (Fiorentini et al., 2010). The progressive cognitive decline and memory loss are strongly associated with neurite outgrowth and synaptic plasticity, which are early and invariant features of AD (Scheff et al., 2007). Compelling evidence also showed that GSK-3β inhibitor could promote neurite outgrowth in SH-SY5Y cells (Lin et al., 2016) and APP/PS1 transgenic mice (Long et al., 2015). Our study showed both VB-030 and VB-037 could benefit the neurite outgrowth against the Tau toxicity in mouse hippocampal primary neurons (Fig. 4), and their effect in reduction of Tau hyperphosphorylation was proved in SH-SY5Y cells (Fig. 3). We suggest the VB-030 and VB-037 are two potent Tau hyperphosphorylation inhibitors with neuroprotective effect for tauopathy, including AD.

Both VB-030 and VB-037 are nitrogen-containing heterocyclic aromatic compounds with quinoline scaffold. Biologically active quinoline derivatives possessing dual effects of inhibition of Aβ toxicity in MC65 neuroblastoma cells and GSK-3β enzyme have been documented (Lu et al., 2014). In our study, quinoline derivatives VB-030 and VB-037 display protective activity against Tau toxicity in mouse hippocampal primary neurons (Fig. 4) and enzymatic inhibitory activity against GSK-3β (Table 2), reinforcing the development of quinoline analogs as anti-AD agents.

In addition to quinoline scaffold, VB-037 also contains a heterocycle morpholine moiety featuring both amine and ether functional groups. Morpholine-containing compounds with potent in vitro GSK-3β inhibitory activity and in vivo Tau phosphorylation inhibitory activity have been addressed (Fukunaga et al., 2013). The potency of VB-037 may be improved by introducing a small alkyl group to the 3-position of the morpholine moiety (Fukunaga et al., 2015). Further development of VB-037 analogs and test of in vivo Tau phosphorylation inhibitory activity are warranted.

In summary, through combinations of virtual screening, GSK-3β enzyme assay, and models of human SH-SY5Y cells expressing Tau folding reporter and mouse hippocampal primary neuron culture under Tau cytotoxicity, two quinoline compounds VB-030 and VB-037 were identified and experimental results showed that they effectively inhibit GSK-3β kinase activity or p-P38 (Thr180/Tyr182) (Fig. 5). Further study will be needed to confirm that this also works in AD animal models and eventually humans.

Figure 5. Model for Tau aggregation reduction and neuronal outgrowth/survival promotion by quinoline compounds VB-030 and VB-037 in AD cell model. The known GSK-3 inhibitor SB-415286 increases the level of p-ERK (Thr202/Tyr204) and p-AKT (Ser473) to increase inactive p-GSK-3β (Ser9), while the tested VB-030 and VB-037 inhibit active GSK-3β and p-P38 (Thr180/Tyr182) respectively, to reduce endogenous Tau phosphorylation at Ser202, Thr231, Ser396 and Ser404 in SH-SY5Y cells, and improved neuronal survival and/or neurite length and branch in mouse hippocampal primary culture under Tau cytotoxicity.

We thank the Molecular Imaging Core Facility of National Taiwan Normal University for the technical assistance. This work was supported by the grants 103-2321-B-182-008, 103-2321-B-003-003, 104-2325-B-003-001 and 104-2325-B-003-003 from the Ministry of Science and Technology, and 104T3040B05 and 104T3040B07 from National Taiwan Normal University, Taipei, Taiwan.

All authors have no conflicts of interest to declare.