Neuroprotective Action of Coumarin Derivatives through Activation of TRKB-CREB-BDNF Pathway and Reduction of Caspase Activity in Neuronal Cells Expressing Pro-Aggregated Tau Protein

Hyperphosphorylation and aggregation of the microtubule binding protein tau is a neuropathological hallmark of Alzheimer’s disease/tauopathies. Tau neurotoxicity provokes alterations in brain-derived neurotrophic factor (BDNF)/tropomycin receptor kinase B (TRKB)/cAMP-response-element binding protein (CREB) signaling to contribute to neurodegeneration. Compounds activating TRKB may therefore provide beneficial effects in tauopathies. LM-031, a coumarin derivative, has demonstrated the potential to improve BDNF signaling in neuronal cells expressing pro-aggregated ΔK280 tau mutant. In this study, we investigated if LM-031 analogous compounds provide neuroprotection effects through interaction with TRKB in SH-SY5Y cells expressing ΔK280 tauRD-DsRed folding reporter. All four LMDS compounds reduced tau aggregation and reactive oxygen species. Among them, LMDS-1 and -2 reduced caspase-1, caspase-6 and caspase-3 activities and promoted neurite outgrowth, and the effect was significantly reversed by knockdown of TRKB. Treatment of ERK inhibitor U0126 or PI3K inhibitor wortmannin decreased p-CREB, BDNF and BCL2 in these cells, implying that the neuroprotective effects of LMDS-1/2 are via activating TRKB downstream ERK, PI3K-AKT and CREB signaling. Furthermore, LMDS-1/2 demonstrated their ability to quench the intrinsic fluorescence of tryptophan residues within the extracellular domain of TRKB, thereby consolidating their interaction with TRKB. Our results suggest that LMDS-1/2 exert neuroprotection through activating TRKB signaling, and shed light on their potential application in therapeutics of Alzheimer’s disease/tauopathies.


Introduction
Neurodegenerative tauopathies, including Alzheimer's disease (AD), frontotemporal dementia, progressive supranuclear palsy and corticobasal syndrome, are characterized by accumulation of misfolded tau proteins [1]. These misfolded tau proteins are hyperphosphorylated and prone to aggregate into insoluble aggregations with a rich β-sheet structure [2]. It has been shown that tau aggregations increase oxidative stress and downregulate the BDNF signaling pathway, eventually leading to neurodegeneration [3,4]. In addition, the misfolded tau protein propagates pathology through connected brain circuits in a prion-like manner [5]. As a result, identification of potential tau misfolding inhibitors is extremely needed to halt neurodegeneration of tauopathies. and LMDS-1 to -4 (1-10 μM) by the thioflavin T assay (n = 3). To normalize, the relative thioflavin T fluorescence of ΔK280 tauRD without compound treatment was set at 100%. Shown below are the EC50 values. (C) Free radical-scavenging activity of kaempferol (as a positive control), LM-031 and LMDS-1 to -4 (10-80 μM) on DPPH (n = 3). Shown below are the EC50 values. (D) Oxygen radical absorbance capacity of LM-031 and LMDS-1 to -4 (1-100 μM) (n = 3). (E) Cytotoxicity of LM-031 and LMDS-1 to -4 against ΔK280 tauRD-DsRed SH-SY5Y cells examined using the MTT assay. Cells were treated with each test compound (0.1-100 μM) and cell viability was measured the next day (n = 3). To normalize, the relative viability of untreated cells was set at 100%.
In addition to activating the TRKB receptor, BDNF also binds to low-affinity nerve growth factor receptor (NGFR) [35], a member of the tumor necrosis factor receptor superfamily. NGFR was expressed in the membrane fractions of SH-SY5Y cells [36]. Increased expression of the NGFR in plasma membrane enhanced mitogen activated protein kinase 8 (JNK) activation and apoptotic cell death in SH-SY5Y cells [37]. We thus examined if these LMDS compounds may act on NGFR. As shown in Figure 6B, LMDS-1/2 treatments did not increase p-JNK (T183/Y185) expression in SH-SY5Y cells expressing ∆K280 tau RD -DsRed (93-108% versus 107%, p > 0.05), suggesting no enhancement of JNK activation through NGFR binding.

Evaluation of TRKB Binding Affinity
The induction of p-TRKB indicated the interaction between LMDS-1/2 and TRKB. Therefore, we cloned into Pichia expression vector pGAPZα A ( Figure 7A) for constitutive expression and purification of recombinant TRKB-ECD-His protein ( Figure 7B). The prepared His-tagged TRKB-ECD protein was used to determine binding specificity of LM-031, LMDS-1 and LMDS-2 via tryptophan fluorescence quenching assay. 7,8-DHF, a selective TRKB agonist with potent neurotrophic activities [38], was included for comparison. The assay relies on the ability to quench the intrinsic protein fluorescence of tryptophan residues which can be selectively measured by exciting at 295 nm. The tryptophan fluorescence differences of TRKB-ECD-His in the presence of test compounds (1-1000 nM) were examined, as binding of 7,8-DHF to TRKB receptor purified from Chinese hamster ovary cells induced a change of the microenvironment of the tryptophan, which was captured by fluorescence spectroscopy [39]. As shown in Figure 7C

Evaluation of TRKB Binding Affinity
The induction of p-TRKB indicated the interaction between LMDS-1/2 and TRKB. Therefore, we cloned into Pichia expression vector pGAPZα A ( Figure 7A) for constitutive expression and purification of recombinant TRKB-ECD-His protein ( Figure 7B). The prepared His-tagged TRKB-ECD protein was used to determine binding specificity of LM-031, LMDS-1 and LMDS-2 via tryptophan fluorescence quenching assay. 7,8-DHF, a selective TRKB agonist with potent neurotrophic activities [38], was included for comparison. The assay relies on the ability to quench the intrinsic protein fluorescence of tryptophan residues which can be selectively measured by exciting at 295 nm. The tryptophan fluorescence differences of TRKB-ECD-His in the presence of test compounds (1-1000 nM) were examined, as binding of 7,8-DHF to TRKB receptor purified from Chinese hamster ovary cells induced a change of the microenvironment of the tryptophan, which was captured by fluorescence spectroscopy [39]. As shown in Figure 7C, tryptophan fluorescence of TRKB-ECD-His was quenched by 7,8-DHF, LM-031, LMDS-1 or LMDS-2 in a concentration-dependent manner, and the fluorescence decrease was maximal at the highest concentration of test compounds. Based on the quantitative analysis of fluorescence change, TRKB-ECD binding affinity (KD) of 7,8-DHF was 16.0 ± 3.4 nM ( Figure 7D). The observed KD of 12.7 ± 2.8 nM, 8.0 ± 17.0 nM and 6.5 ± 6.6 nM of LM-031, LMDS-1 and LMDS-2 demonstrated the high binding affinities to the ECD of the TRKB receptor.

Discussion
The pathogenesis of tauopathies is still unclear, while the treatment to halt taumediated neurodegeneration remains unavailable. Here, we demonstrate that overexpression of pro-aggregated ∆K280 tau RD in SH-SY5Ycells leads to increased aggregation and oxidative stress, neurite outgrowth defects, and up-regulation of caspase-1 and caspase-6 activities. ∆K280 tau RD also down-regulates the phosphorylation of TRKB and CREB, as well as the expression of CREB, BDNF and BCL2, and increases BAX expression. Administrations with coumarin derivatives LMDS-1/2 rescue these neurodegenerative phenotypes by increasing TRKB phosphorylation and BDNF expression. Knockdown of TRKB and treatments with ERK inhibitor U0126 or PI3K inhibitor wortmannin counteract the neuroprotective effects of LMDS-1/2. The tryptophan fluorescence quenching assay further confirms the direct interactions between LMDS-1/2 and the TRKB extracellular domain. These results shed light on the role of TRKB signaling in tau-mediated neurodegeneration. The neuroprotective effects of LMDS-1/2 further suggest the potential of TRKB agonists in treating tauopathies.
A few TRKB agonists have been examined in cell or animal models for neurodegenerative diseases. For example, 7,8-DHF has the potential to attenuate Aβ deposition, loss of hippocampal synapses and memory deficits in transgenic AD mice [40,41]. 7,8-DHF also reduces ROS production in HT-22 hippocampal neuronal cells [42], and protects PC12 pheochromocytoma cells against 6-hydroxydopa-induced cell death [43]. Intranasal administration with LM22A-4 activates TRKB to improve motor learning after traumatic brain injury in rats [44]. However, the limited ability for crossing the BBB restricts the application of LM22A-4 in treating neurodegenerative diseases [44]. In our study, LMDS-1/2 directly bind to TRKB, elicit downstream signaling, including ERK and PI3K-AKT pathways, and protect neurons against tau-mediated neurotoxicity. In the tryptophan fluorescence quenching assay, LMDS-1/2 demonstrate adequate binding affinity to TRKB. Both compounds reduce ROS and improve neurite outgrowth in ∆K280 tau RD -DsRed SH-SY5Y cells. Blockage of either ERK or Pl3K-AKT signaling counteracts the improvement of neurite outgrowth by LMDS-1/2, suggesting activations of both signals are necessary for neuroprotection.
Highly expressed in neural tissues, ERK plays an important role in the survival of neurons [45]. Transgenic activation of ERK in mice reduces the 1-methyl-4-phenyl-1,2,3,6tetrahydropyridine (MPTP)-or facial nerve axotomy-induced loss of dopaminergic or motor neurons [46]. On the other hand, a number of reports have shown a relationship between neurodegeneration and increased ERK activity. Activation of ERK has been found in hippocampal slides from APP-transgenic AD mice [47]. ERK is colocalized with tau and neurofibrillary tangles in neurons [48]. The phosphorylation of ERK is elevated in brain extracts from AD patients [49]. Tau is also a substrate of phosphorylation by ERK [50]. Importantly, ERK signaling can be activated by c-reactive protein or complements [51,52], both of which are colocalized with neurofibrillary tangles in brains of AD patients [52][53][54]. The complex role of ERK activation and its interactions with other signaling pathways in neurodegeneration need to be elucidated.
The PI3K-AKT signaling pathway plays an important role in neuronal survival, proliferation, differentiation, autophagy and neurogenesis [55]. The activation of the PI3K-AKT signal pathway is able to protect neurons against tau aggregation. In AD, the activation of glycogen synthase kinase-3β (GSK-3β) is closely related to the hyperphosphorylation and accumulation of tau [56]. GSK-3β is rendered inactive when it is phosphorylated at S9 by AKT [57]. The postmortem study showed decreased levels of PI3K subunits and reduced phosphorylation of AKT in the brains of AD patients [58,59]. Administering AKT inhibitor wortmannin to rat lateral ventricles increases tau hyper-phosphorylation and causes axonal swelling [60]. In our study, inhibition of AKT by wortmannin attenuates the rescue of neurite outgrowth by LMDS-1/2, suggesting that the PI3K-AKT signaling pathway is mandatory for axon growth. On the other hand, tau is also a phosphorylation substrate of AKT, while AKT inhibition leads to the decrease in tau phosphorylation at T212 and S214 [61,62]. Importantly, TRKB-PI3K-AKT signaling also promotes the activation of NRF2 antioxidant signaling [63], while disruption of antioxidant signaling increases the production of ROS and further impairs cognitive function in animal models [64][65][66][67]. The net effect of PI3K-AKT inhibition in tauopathies should be further investigated.
The activations of ERK and PI3K-AKT by LMDS-1/2 further phosphorylate CREB. CREB is a critical nuclear transcription factor needed for neuron survival [22]. Phosphorylated CREB binds to the coactivators CREB binding protein (CBP) and E1A binding protein p300 (EP300) to up-regulate the expression of target genes such as BDNF and BCL2 [20,68]. It is of interest that the promoter of tau gene MAPT contains CRE-like elements and overexpression of CREB down-regulates the expression of tau in SH-SY5Y cells [69]. Our results show that application of either U0126 or wortmannin down-regulates CREB phosphorylation and expression of BDNF and BCL2, and up-regulates BAX expression, supporting the regulation of CREB activity by ERK and PI3K-AKT signaling pathways. The neuroprotection of LMDS-1/2 against tau-mediated neurotoxicity is also counteracted by the treatment of both inhibitors, implicating the contribution of ERK, PI3K-AKT and CREB signaling pathways to the neuroprotection in the ∆K280 tau RD cell model. Furthermore, our group also demonstrated that LMDS-1 could up-regulate the TRKB-ERK-CREB pathway and BDNF expression in hippocampal primary neurons incubated with oligomeric Aβ and Aβ-induced AD mice [70]. These results support the potential of coumarin derivatives in improving neurodegeneration.
The fluorescent tryptophan quenching assay supports that LMDS-1 and LMDS-2 directly interact with TRKB-ECD. The K D of these two compounds are lower compared to 7,8-DHF and LM-031, suggesting their high affinities to TRKB-ECD, whereas their binding sites in TRKB-ECD remain elusive. BDNF interacts with the leucine-rich repeat (LRR) motif and Ig2 domain of TRKB-ECD [71]. Molecular modeling supports that 7,8-DHF could insert itself into to the pocket between the N-terminal cap and the first repeat of the LRR domain [72]. However, TRKB-ECD is a highly glycosylated protein that contains 10 N-linked glycosylation sites [73]. Six disulfide linkages are also formed within two cysteine cluster domains in the N-terminus of TRKB [73]. It remains uncertain whether Pichia pastoris-derived recombinant proteins could recapitulate these posttranslational modifications. Further studies, such as the cocrystal structure analysis, would provide more spatial details on the direct interaction between LMDS1/2 and TRKB. Apart from binding to TRKB, LMDS-1/2 also directly reduce the aggregation of E. coli-derived ∆K280 tau RD in thioflavin T assay. It is also possible that the reduction of tau aggregation can modulate ERK, PI3K-AKT and CREB signaling pathways. Alternatively, LMDS-1/2 may also activate other receptors to modulate the above signaling pathways and their neuroprotective effects. Further studies will be warranted to clarify whether ERK, PI3K-AKT and CREB are up-regulated directly by TRKB activation, by aggregate inhibition, or by a pleiotropic mechanism through activation of other receptors. In vivo experimentation is also needed to confirm these promising results.

Coumarin Compounds
Coumarin derivatives LMDS-1 to -4 were purchased from Enamine (Kyiv, Ukraine). In-house LM-031 activating the CREB-dependent survival and anti-apoptosis pathway [26] was included for comparison. In addition, Congo red and kaempferol were obtained from Sigma-Aldrich Co. (St. Louis, MO, USA) as positive controls for monitoring tau folding and free radical-scavenging activity, respectively.

Anti-∆K280 Tau RD Aggregation Assay
Thioflavin T is widely used to visualize and quantify the presence of misfolded amyloid protein aggregates in vitro [74]. The tau aggregation inhibiting potential of Congo red, LM-031 and LMDS-1 to -4 was assessed by using E. coli-expressed ∆K280 tau RD protein [28]. Briefly, purified ∆K280 tau RD protein (20 µM) was incubated with Congo red or coumarin compounds (1-10 µM) in 150 mM NaCl and 20 mM Tris-HCl pH8.0, at 37 • C with shaking for 48 h to form aggregates. Then, thioflavin T (final 5 µM; Acros Organics, Geel, Belgium) was added and incubated for 25 min at room temperature. Fluorescence intensity of samples was recorded at excitation/emission wavelengths of 420/485 nm using an FLx800 microplate reader (BioTek, Winooski, VT, USA). The half maximal effective concentration (EC 50 ) values were calculated by using a linear interpolation method.

Antioxidant Assay
The potential free radical-scavenging activity of test compounds was assayed using stable free radical 1,1-diphenyl-2-picrylhydrazyl (DPPH) (Sigma-Aldrich). Briefly, kaempferol, LM-031 or LMDS-1 to -4 (10-80 µM) was added to DPPH (100 µM in 99% ethanol). The solution was vortexed for 15 s, left for 30 min at room temperature, the mixture was transferred to a 96-well UV-transparent microplate and the absorbance at 517 nm was measured using a Multiskan TM GO microplate spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Radical-scavenging activity was calculated using the equation: 1 − (absorbance of sample/absorbance of control) × 100%, and EC 50 estimated using the linear interpolation method.
In addition, an oxygen radical antioxidant capacity (ORAC) assay was performed using an OxiSelect™ kit according to the manufacturer's instruction (Cell Biolabs, San Diego, CA, USA). Briefly, Trolox standards (2.5-50 µM) and samples (4-100 µM) were diluted with 50% acetone. Blank (50% acetone), standards and samples were mixed with fluorescein and incubated at 37 • C for 30 min. Following the free radical initiator 2,2azobis(2-methylpropionamidine) dihydrochloride (AAPH), the produced peroxyl radicals (ROO•) quench the fluorescent probe over time [75]. Antioxidants present in the assay block the peroxyl radical oxidation of the fluorescent probe. The course of the reaction was recorded for 60 min, with one measurement every five minutes using a BioTek FLx800 microplate reader. The excitation and emission wavelengths were set at 480 nm and 520 nm, respectively. To quantify the oxygen radical antioxidant activity in a sample, the area under the curve (AUC) for blank, standard and samples was calculated. After subtraction of the blank, the equivalent Trolox concentrations of samples were expressed based on the Trolox standard curve.

Real-Time PCR Analysis
Total RNA of ∆K280 tau RD -DsRed SH-SY5Y cells on day 8 was extracted using Trizol reagent (Sigma-Aldrich) and reverse transcribed using SuperScript TM III reverse transcriptase (Invitrogen). Quantitative PCR was performed with 50 ng cDNA and customized Assays-by-Design probe for DsRed [12] and HPRT1 (4326321E, endogenous control) in a 96-well real-time PCR instrument (StepOnePlus TM Real-time PCR system; Applied Biosystems, Foster City, CA, USA). Fold difference of RNA was calculated using the formula 2 ∆Ct , ∆C T = C T (HPRT1) − C T (DsRed), in which C T indicates cycle threshold.

High-Content Analysis of Neurite Outgrowth
As described, ∆K280 tau RD -DsRed SH-SY5Y cells were seeded in a 24-well plate (5 × 10 4 /well) with retinoic acid addition on day 1, treated with tested compounds (10 µM) and ∆K280 tau RD -DsRed expression was induced with doxycycline (2 µg/mL) on day 2. On day 8, after being fixed in 4% paraformaldehyde for 15 min, permeabilized in 0.1% Triton X-100 for 10 min and blocked in 3% bovine serum albumin (BSA) for 20 min, the cells were stained with TUBB3 (neuronal class III β-tubulin) primary antibody (1:1000; BioLegend, San Diego, CA, USA) at 4 • C overnight, followed by goat anti-rabbit Alexa Fluor ® 555 secondary antibody (1:1000; Molecular probes) at room temperature for 2 h, with 4 -6-diamidino-2phenylindole (DAPI, 0.1 µg/mL; Sigma-Aldrich) included for nuclei staining. Neuronal images were captured using the high-content analysis system as described. Neurite total length (µm), processes (primary neurite extensions projecting directly from the cell body) and branching (points at which primary neurites bifurcated) were analyzed using Neurite Outgrowth Application Module (MetaXpress; Molecular Devices). Around 5000 cells were analyzed in each of three independent experiments for each sample.

Caspase-1, -3 and -6 Activity Assays
∆K280 tau RD -DsRed SH-SY5Y cells were seeded in a 6-well plate (5 × 10 5 /well) and treated with retinoic acid, test compound and doxycycline as described. On day 8, cells were collected and resuspended in lysis buffer, followed by six freeze/thaw cycles and centrifugation to collect the supernatant. Caspase-1 activity in cell extracts was measured using a caspase-1 fluorometric assay kit (YVAD-AFC substrate; BioVision, Milpitas, CA, USA). In addition, a caspase-6 fluorometric assay kit (VEID-AFC substrate; BioVision) was used to measure caspase-6 activity. The mixture was incubated for 2 h at 37 • C. A FLx800 microplate reader (BioTek) was used to measure caspase-1 or caspase-6 activity with excitation and emission wavelengths of 400 nm and 505 nm, respectively. For caspase-3 activity measurement, a caspase-3 fluorometric assay kit (DEVD-AMC substrate, Sigma-Aldrich) was used. The excitation and emission wavelengths of AMC were 360 nm and 460 nm, respectively.

RNA Interference
Lentiviral short hairpin RNA (shRNA) targeting TRKB (TRCN0000002243, TRCN0000002245 and TRCN0000002246) and a negative control scrambled (TRC2.Void, ASN000001) [77] from National RNAi Core Facility, IMB/GRC, Academia Sinica, Taipei, Taiwan were applied to knock down TRKB expression in ∆K280 tau RD -DsRed SH-SY5Y cells. On day 1, cells were plated on 6-well (for protein analysis) or 24-well (for neurite outgrowth analysis) plates in the presence of retinoic acid as described. On day 2, the cells were infected with lentivirus (multiplicity of infection of 3 for each shRNA) in medium containing polybrene (8 µg/mL; Sigma-Aldrich) to increase the efficiency of viral infection. On day 3, the culture medium was changed and the cells were pretreated with LM-031 or LMDS-1 to -4 (10 µM) for 8 h, followed by inducing ∆K280 tau RD -DsRed expression. On day 9, the cells were collected for TRKB protein analysis or analyzed for neurite outgrowth as described.

TRKB-ECD Protein Preparation
The cDNA fragment containing the full-length extracellular domain of TRKB (TRKB-ECD) (residues 32-429, XP_016870240) was amplified from pDONR223-NTRK2 plasmid (Addgene, Watertown, MA, USA) using forward primer 5 -ctcgagaaaagagaggctgaagctTGTC CCACGTCCTGAAATG (XhoI site and Kex2 cleavage of the signal sequence in lowercase letters) and reverse primer 5 -gtcgacTTCCCGACCGGTTTTATCAGTG (SalI site in lowercase letters). After cloning and sequencing, the TRKB-ECD fragment was excised with XhoI and SalI and subcloned into Pichia expression vector pGAPZα A (Invitrogen). The AvrII-linearized plasmid was then transformed into Pichia pastoris, and stable transformants selected and cultured at 30 • C for 4 days for constitutive expression of recombinant His-tagged TRKB-ECD. The secreted TRKB-ECD-His in cultured medium was purified using Ni-NTA His•Bind resins (Novagen, Madison, WI 53719, USA) according to the supplier's instructions. Protein purity was examined by Coomassie blue staining of SDS-polyacrylamide gel as well as immunoblotting using anti-His (1:2000; Acris, Herford, Germany) and anti-TRKB (1:1000; Cell Signaling) antibodies.

Tryptophan Fluorescence Quenching Assay
The interaction between TRKB-ECD-His and LM-031, LMDS-1 or LMDS-2 was determined by tryptophan fluorescence titration. 7,8-DHF (Sigma-Aldrich) was included as a positive control [39]. Tryptophan fluorescence spectra of TRKB-ECD-His (50 nM in PBS) were obtained before and after titration with different concentrations of test compounds (0-1000 nM). The fluorescent quenching was conducted in a cuvette with a 1 cm path length cell by an F-7000 fluorescence spectrophotometer (Hitachi, Tokyo, Japan) at room temperature. The intrinsic fluorescence spectra of TRKB were recorded with excitation wavelength set at 295 nm and emission wavelength in the range 300-400 nm. The dissociation constant (K D ) of 7,8-DHF, LM-031, LMDS-1 and LMDS-2 to TRKB was determined by fitting the normalized fluorescence intensity at 330 nm under different concentrations of test compounds using the equation derived from Liu et al. [39]: where f is the fractional change, and [P] T and [D] T are the total concentration of TRKB-ECD-His protein and test compound, respectively.

Statistical Analysis
The presented data are shown as mean ± SD of three independent experiments in two or three biological replicates. To compare the differences between groups, two-tailed Student's t-test or one-way analysis of variance (ANOVA) was performed, with Tukey's post hoc test where appropriate. p values < 0.05 were considered as statistically significant.

Conclusions
In the present study, we show that the TRKB signaling pathway in ∆K280 tau RD SH-SY5Y cells is compromised, resulting in BCL2 and BDNF down-regulations, BAX upregulation, ROS overproduction and neurite outgrowth impairments. Binding to TRKB, activating its downstream ERK, PI3K-AKT and CREB signaling pathways and reducing caspase activity, LMDS-1/2 exert neuroprotective effects in these cells. Blockage of either ERK or PI3K-AKT signaling pathways counteracted their neuroprotection against tau aggregation, indicating both signaling pathways are indispensable to protect neurons against tau-mediated neurotoxicity. Given that the effects of LMDS-1/2 were only examined in cell models, future studies in animal models of tauopathies or other neurodegenerative diseases will be necessary to validate their potential in treating patients with AD and tauopathies.