Novel TRKB agonists activate TRKB and downstream ERK and AKT signaling to protect Aβ-GFP SH-SY5Y cells against Aβ toxicity

Decreased BDNF and impaired TRKB signaling contribute to neurodegeneration in Alzheimer’s disease (AD). We have shown previously that coumarin derivative LM-031 enhanced CREB/BDNF/BCL2 pathway. In this study we explored if LM-031 analogs LMDS-1 to -4 may act as TRKB agonists to protect SH-SY5Y cells against Aβ toxicity. By docking computation for binding with TRKB using 7,8-DHF as a control, all four LMDS compounds displayed potential of binding to domain d5 of TRKB. In addition, all four LMDS compounds exhibited anti-aggregation and neuroprotective efficacy on SH-SY5Y cells with induced Aβ-GFP expression. Knock-down of TRKB significantly attenuated TRKB downstream signaling and the neurite outgrowth-promoting effects of these LMDS compounds. Among them, LMDS-1 and -2 were further examined for TRKB signaling. Treatment of ERK inhibitor U0126 or PI3K inhibitor wortmannin decreased p-CREB, BDNF and BCL2 in Aβ-GFP cells, implicating the neuroprotective effects are via activating TRKB downstream ERK, PI3K-AKT and CREB signaling. LMDS-1 and -2 are blood–brain barrier permeable as shown by parallel artificial membrane permeability assay. Our results demonstrate how LMDS-1 and -2 are likely to work as TRKB agonists to exert neuroprotection in Aβ cells, which may shed light on the potential application in therapeutics of AD.

brain and plays a vital role in promoting neural plasticity, neuronal growth and cell survival [9]. BDNF has a high binding affinity to tropomyosinrelated kinase B (TRKB), also known as neurotrophic receptor tyrosine kinase 2, to induce TRKB dimerization and phosphorylation of TRKB, and subsequently to activate downstream cascades including, phosphoinositide 3-kinase (PI3K)-AKT serine/threonine kinase (AKT), extracellular signalregulated kinase (ERK)-cAMP responsive element binding protein 1 (CREB), and phospholipase C-γ1 (PLC-γ1). The TRKB receptor activation would lead to enhanced memory formation and storage as well as neuroplasticity, neurogenesis, neurite outgrowth and neuronal survival [10,11]. ERK phosphorylates CREB to promote transcription of genes for neuronal survival, neurite outgrowth and neuroplasticity [12], such as BDNF [13] and BCL2 apoptosis regulator (BCL2) [14]. The post-synaptic ERK signal would also enhance the neuroplasticity related long-term potentiation [11].
Decreased BDNF has been shown in several neurodegenerative diseases [12]. Previous reports have shown that BDNF transcription and protein expression is reduced in hippocampus, cortex and Meynert basal ganglion of AD brains [13,14]. It has been shown that accumulation of Aβ is associated with loss of BDNF [15] and Aβ inhibits protein kinase A to down-regulate the CREB phosphorylation and expression of BDNF [15,16]. Oligomeric Aβ also interferes with Ras-ERK and PI3K-AKT pathways to aggravate neurotoxicity [17]. Several preclinical studies using AD mouse models have demonstrated that increased BDNF expression rescues neurotoxicity and improves cognitive function [18][19][20]. Nevertheless, the poor bioavailability of BDNF, or example the short half-life in plasma and the limited blood-brain barrier (BBB) permeability [21], restricts the application of BDNF.
Developing novel small molecule agonists of TRKB receptor is a potential therapeutic strategy for neurodegenerative diseases including AD [22]. To identify compounds acting as TRKB agonists with a good BBB permeability is crucial for developing therapeutics for AD. Our previous studies have demonstrated that a novel coumarin derivative, LM-031, displayed neuroprotective effects by upregulating CREB/BDNF/BCL2 pathway in our Flp-In Aβ-GFP SH-SY5Y cells [23]. In this study, we searched three online databases for LM-031 analogous compounds using compound similarity search software tools and identified four top-scoring compounds LMDS-1 to -4 as the candidate TRKB agonists. To examine their plausible role as TRKB agonists, docking computation was firstly conducted to investigate binding strength and conformation of LM-031 and the four analogs. We then examined if the four potential small TRKB agonists would exert neuroprotective effects and influence the TRKB downstream pathways in Aβ-GFP folding reporter SH-SY5Y cells. The BBB permeability was also examined by using parallel artificial membrane permeability assay (PAMPA).

Test compounds
We tested the coumarin derivative LM-031 and the analogs LMDS-1 to -4 ( Figure 1A). According to their molecular weight (MW), hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), and octanol-water partition coefficient (cLogP), these five compounds fulfill five criteria of Lipinski's rule for predicting oral bioavailability [24] ( Figure 1B). All five compounds were anticipated to diffuse over the BBB with a polar surface area (PSA) of smaller than 90 Å 2 [25], which were similarly supported by an online BBB predictor [26] ( Figure 1B).
Anti-amyloid and anti-oxidative stress are regarded as crucial AD therapeutic approaches. The thioflavin T fluorescence assay was applied to gauge the suppression of Aβ aggregation. Curcumin is known to reduce amyloid aggregation [27]. Both curcumin and coumarin were included for comparison. Half maximal effective concentration (EC50) values of curcumin, LM-031, LMDS-1 to -4 and coumarin to inhibit Aβ aggregation were: <5, 9, <5, 18, 14, <5 and 25 μM, respectively ( Figure 1C). The free radical scavenging activity of the LM-031, analogs and coumarin was investigated using diphenylpicrylhydrazyl (DPPH) as a substrate. Kaempferol (a positive control [28]), LM-031, LMDS-1 to -4 and coumarin had EC50 values of 28,104,117,146,167,116 and 179 μM, respectively ( Figure 1D). Figure 2A shows the binding conformations of 7,8-DHF, LM-031 and two representative LMDS compounds (LMDS-1 and -4) and Figure 2B shows the predicted binding strengths of these compounds. Of the tested compounds, the computations predicted that LMDS-1 and -2 were the top two compounds interacting with TRKB receptor. Note that the LMDS compounds have different binding modes in comparison with 7,8-DHF and LM-031. In addition, interestingly, all LMDS-1 to -3 have the same binding mode, 3 hydrogen bonds with backbone of L315 and I334, and side chain of K312. In contrast, LMDS-4 binds with d5 domain in different orientation with a lower binding strength compared with LMDS-1 to -3. It only has a bifurcated hydrogen bond with side chain of K312. These results suggest that bulky carboxylate group is not favorable functional group at the position of fused ring of LMDS compounds shown in Figure 1A.

Aβ aggregation inhibition and reduction of Aβinduced oxidative stress
Aβ-GFP SH-SY5Y cells [29] were used to examine Aβ aggregation inhibition of LM-031 and analogs ( Figure 3A). The cell model uses GFP as a reporter to reflect the level of Aβ misfolding. The fast misfolding and formation of Aβ aggregates causes misfolding of fused GFP, thereby suppressing green fluorescence. Compounds that inhibit Aβ misfolding allow refolding of GFP monitored by an increase in green fluorescence on Aβ-GFP expressing cells [30]. Curcumin, which can modify the Aβ aggregation pathway and ameliorate Aβinduced toxicity [31], was considered as a positive control. Coumarin was also included for comparison.

Comparative effects of LMDS-1, -2 and BDNF on TRKB signaling
As LMDS-1 and -2 target ERK, AKT and downstream CREB, effects of BDNF and LMDS compounds on TRKB signaling were compared. Oxidative stress attenuation was first evaluated by treating Aβ-GFP cells with BDNF at 1-100 ng/ml concentration or LMDS-1, -2 at 1.2-5 μM concentration ( Figure 7A). The ROS level induced by Aβ overexpression was successfully

BBB permeability of LM-031 and analogs
As of now, the PAMPA model has been proposed to analyze membrane permeability, including BBB [32,33]. The in vitro BBB permeability of LM-031 and analogs was then estimated using the PAMPA-BBB approach. Quality controls, including high permeability bupropion [33], low permeability piroxicam [32], and integrity marker lucifer yellow, were included for comparison. Bupropion and piroxicam, which represent high (> 4 × 10 −6 cm/s) and low (< 2 × 10 −6 cm/s) BBB permeable controls, had effective permeability (Pe) values of 17.06 and 1.65 (10 −6 cm/s), respectively (

DISCUSSION
There are several reasons for identifying small molecules acting as novel potential TRKB agonists for AD therapeutic uses. First, there is no available treatment currently to cure AD or to halt its progression. Second, substantial evidence has shown that decreased BDNF and impaired TRKB pathways, including ERK, CREB and PI3K-AKT, contribute to neurodegeneration in AD [13][14][15][16][17]. Third, agents enhancing BDNF expression or TRKB agonists provide beneficial effects to AD [35][36][37]. Fourth, BDNF hardly penetrates BBB into brain [21]. Thus, identifying TRKB agonists with good BBB permeability may provide a therapeutic strategy for AD. In the present study, we identified novel potential TRKB agonists with good bioavailability (Figure 1). This is supported by the calculated values of MW, HBD, HBA and cLogP of these compounds because they satisfy the Lipinski's criteria [24]. Furthermore, PSA of less than 90 Å 2 [25] and predicted BBB scores of greater than threshold 0.02 [26] support a good BBB permeability of these compounds (Figure 1). The binding scores predicted by docking computations suggest these compounds have a good affinity to d5-domain of TRKB (Figure 2), providing evidence of their potential as TRKB agonists.
In addition to targeting TRKB, neurotrophic receptor tyrosine kinase 1 (TRKA) agonist D3 have been shown to provide beneficial effects to activate TRKA-related signaling cascades and enhance cholinergic neurotransmission in transgenic mice overexpressing human APP with KM670/671NL and V717F mutations [42]. Targeting neurotrophic receptor tyrosine kinase 3 (TRKC) signaling may also be an effective approach to rescue impaired cholinergic function in AD [43]. Our study results demonstrate that LMDS compounds specific for TRKB binding enhance neurite outgrowth by activating signaling downstream of TRKB. Whether TRK receptor isoforms TRKA and TRKC could potentially be activated by LMDS compounds remains to be determined. In addition, future in vitro biding assay should be performed to provide evidence of LMDS compounds binding to extracellular domain of TRKB [22] to show their specificity of TRKB binding.

CONCLUSIONS
In this study, we identified LM-031 analogous compounds LMDS-1 and -2 which may serve as the potential TRKB agonists to treat AD. This is supported by our experimental results demonstrating TRKB signaling-activating, aggregation-inhibitory, oxidative stress-and caspase 1-reducing, and neurite outgrowthpromoting effects in Aβ-GFP SH-SY5Y cells. Pursuit of analogous compounds resembled to the identified compounds should be of interest in the future.

Docking computation
Docking calculations for LM-031 and analogs were carried out using the GOLD docking program [62,63]. The computational protocol was similar to our previous study [64]. Briefly, protein structure of domain d5 of TRKB (pdb code: 1HCF) [65] was used for docking compounds. The d5 domain determined the specificity of neurotrophin receptors experimentally [66]. Binding site of small molecule agonists has been predicted by a previous computation [67]. To assure convergence of the computed findings, the numbers of operations of 10,000, 20,000, 40,000 and 80,000 were carried out in the current docking computations of LM-031 and analogs. For comparison, 7,8-DHF, a bioactive TRKB agonist [22], was also added to the computation.

Real-time PCR analysis
Total RNA was extracted, converted to cDNA, and the expressed Aβ-GFP RNA was quantified as stated [69]. The formula 2 ΔCt , ΔCT = CT (HPRT1) -CT (EGFP), where CT stands for cycle threshold, was used to compute fold change.

Caspase 1 and AChE activity assays
Cells on 6-well plate (5 × 10 5 /well) were subjected to the retinoic acid, test compound and doxycycline treatments as stated. Cells were collected on day 8 and cell lysates prepared for caspase 1 (BioVision, Milpitas, CA, USA) and AChE (Sigma-Aldrich) activity measurement [60].

TRKB RNA interference
To knock down TRKB expression in Aβ-GFP SH-SY5Y cells, lentiviral short hairpin RNA (shRNA) targeting TRKB and control (scrambled) were used [69]. Cells were plated in the presence of retinoic acid (5 × 10 5 /6-well for protein analysis or 6 × 10 4 /24-well for neurite outgrowth analysis) on day 1, infected with lentivirus on day 2, pretreated with test compound (5 µM) plus inducing Aβ-GFP expression on day 3, and collected for neurite outgrowth or TRKB protein analysis on day 9 as stated [69]. AGING

PAMPA to assess BBB permeability
The permeability of LMDS-1 and -2 was determined by a PAMPA-BBB assay. Firstly, 300 μl of LMDS-1 or -2 (1 μM) solution were added to the donor well (Millipore). Bupropion (a high permeability marker), piroxicam (a low permeability marker), and lucifer yellow (an integrity marker) (Sigma-Aldrich) were used as quality control (QC) compounds for comparison. The sandwich with aqueous donor on the bottom, artificial lipid membrane (PVDF filter coated with porcine polar brain lipid) in the middle, and aqueous acceptor (5% DMSO in PBS) on the top was assembled as stated [69]. Each compound was tested in triplicate. After incubation for 18 h at the ambient temperature, the PAMPA sandwich plate was separated. The concentrations of test and QC compounds in the donor and acceptor wells were measured as stated [69], and the effective permeability coefficient (Pe) was computed [70].

Statistical analysis
The mean ± standard deviation from three independent experiments are reported as the data. As appropriate, a two-tailed Student's t test or one-way ANOVA (analysis of variance) with a post hoc Tukey test were used to assess group differences. P values < 0.05 indicated statistical significance.