Kinase Inhibitors with Antiepileptic Properties Identified with a Novel in Vitro Screening Platform

Kinase signaling plays an important role in acquired epilepsy, but only a small percentage of the total kinome has been investigated in this context. A major roadblock that prevents the systematic investigation of the contributions of kinase signaling networks is the slow speed of experiments designed to test the chronic effects of target inhibition in epilepsy models. We developed a novel in vitro screening platform based on microwire recordings from an organotypic hippocampal culture model of acquired epilepsy. This platform enables the direct, parallel determination of the effects of compounds on spontaneous epileptiform activity. The platform also enables repeated recordings from the same culture over two-week long experiments. We screened 45 kinase inhibitors and quantified their effects on seizure duration, the frequency of paroxysmal activity, and electrographic load. We identified several inhibitors with previously unknown antiepileptic properties. We also used kinase inhibition profile cross-referencing to identify kinases that are inhibited by seizure-suppressing compounds, but not by compounds that had no effect on seizures.


Introduction
Characteristics of the epileptic brain include axon sprouting, synaptic reorganization, inflammation, and hyperexcitability [1,2]. Axon and dendritic growth and development, synaptic formation and maintenance, changes in receptor number and composition, and inflammatory processes are regulated by cell signaling pathway networks [2][3][4][5][6][7][8][9]. The inhibition of these pathways has the potential to prevent the formation of epileptic circuitry and thus prevent epilepsy development after brain injury, or even disrupt existing epileptic circuits and have a permanent disease-modifying effect. In contrast, existing anticonvulsants do not prevent or cure epilepsy.
Kinase signaling, including activation of PI3K-Akt-mTOR, JAK-STAT, and BDNF-TrkB pathways, has been implicated in animal and in vitro models of acquired chronic epilepsies [10][11][12][13][14][15][16][17]. However, kinases that have been found to play a role in epilepsy to date represent only a small percentage of the total kinome, since the human genome includes over 500 kinase genes [18]. Many of these kinases play important roles in neurons, glia, and microglia [19], and there is a significant likelihood that they may be involved in epileptogenesis and/or the progression of epilepsy. Small molecule inhibitors have been synthesized for many of these kinases, and used extensively in cancer research, with some inhibitors successfully passing clinical trials [20]. While kinase inhibitors are not always specific, their inhibition profiles have been experimentally determined and described [21][22][23]. Despite the availability of inhibitors, only a few kinases have been investigated with the goal of preventing or modifying epilepsy. For example, receptor tyrosine kinases (RTKs), a class of kinases that act as cell surface receptors, include FGFRs, VEGFRs, Flt, EGFR, Erbb receptors, IGF-1R, c-Met, cFMS, GM-CSFR, and PDGFRs as well as neurotrophin receptors (TrkB and others). All of the RTKs listed above (and their ligands) are expressed in the brain, and changes in their expression or phosphorylation have been reported in the injured or epileptic brain [24][25][26][27][28][29][30][31]. However, only a few non-neurotrophin RTKs, such as IGF-1R, have been explored as potential targets for antiepileptic drugs [32]. The RTK signaling network is highly complex. At least 20 RTKs are expressed in the hippocampus at significant levels [33], and their downstream signaling includes multiple pathways such as PI3K-Akt-mTOR, multiple MAPKs, Stat, and Rho-ROCK ( Figure 1) [34][35][36][37][38][39][40][41]. Downstream kinases, with the exception of those involved in mTOR signaling, have also been under-explored as antiepileptic drug targets. likelihood that they may be involved in epileptogenesis and/or the progression of epilepsy. Small molecule inhibitors have been synthesized for many of these kinases, and used extensively in cancer research, with some inhibitors successfully passing clinical trials [20]. While kinase inhibitors are not always specific, their inhibition profiles have been experimentally determined and described [21][22][23].
Despite the availability of inhibitors, only a few kinases have been investigated with the goal of preventing or modifying epilepsy. For example, receptor tyrosine kinases (RTKs), a class of kinases that act as cell surface receptors, include FGFRs, VEGFRs, Flt, EGFR, Erbb receptors, IGF-1R, c-Met, cFMS, GM-CSFR, and PDGFRs as well as neurotrophin receptors (TrkB and others). All of the RTKs listed above (and their ligands) are expressed in the brain, and changes in their expression or phosphorylation have been reported in the injured or epileptic brain [24][25][26][27][28][29][30][31]. However, only a few non-neurotrophin RTKs, such as IGF-1R, have been explored as potential targets for antiepileptic drugs [32]. The RTK signaling network is highly complex. At least 20 RTKs are expressed in the hippocampus at significant levels [33], and their downstream signaling includes multiple pathways such as PI3K-Akt-mTOR, multiple MAPKs, Stat, and Rho-ROCK ( Figure 1) [34][35][36][37][38][39][40][41]. Downstream kinases, with the exception of those involved in mTOR signaling, have also been under-explored as antiepileptic drug targets. in the hippocampus, based on Allen Brain Atlas mouse data (www.brain-map.org). Most RTKs were concentrated in the pyramidal layers of the hippocampus with the exception of PDGFRα, which had strong but diffuse expression; (B) Downstream signaling activated by RTKs. Kinases that were inhibited in this work are highlighted in red. Cross-talk with JAK/STAT and Wnt signaling is also shown.
A major roadblock that prevents the systematic investigation of the contributions of RTKs and their signaling network to epilepsy is the slow speed of experiments designed to test the chronic effects of target inhibition in animal models of epilepsy. We have recently reported a novel in vitro antiepileptic drug screening platform that integrated an organotypic hippocampal culture model of epilepsy with microfluidic-multiple electrode array (μflow-MEA) chips [42]. The platform was capable of monitoring chronic epileptiform activity in inhibitor-treated cultures. A pilot screen conducted with the μflow-MEA platform revealed antiepileptic effects of EGFR/ErbB-2 and cFMS inhibitors. However, the transferability of the platform to other laboratories was limited due to complex clean room microfabrication requirements of the μflow-MEA chips. Here, we report a simplified screening platform that is based on standard tissue culture plates and commercially available microwire electrodes. We used this platform to evaluate the effects of 45 inhibitors of RTKs and their signaling pathways (Table 1) on the epileptiform activity in organotypic hippocampal cultures. We then used the data on kinase inhibitor selectivity to determine the likely kinase targets of the efficacious inhibitors. in the hippocampus, based on Allen Brain Atlas mouse data (www.brain-map.org). Most RTKs were concentrated in the pyramidal layers of the hippocampus with the exception of PDGFRα, which had strong but diffuse expression; (B) Downstream signaling activated by RTKs. Kinases that were inhibited in this work are highlighted in red. Cross-talk with JAK/STAT and Wnt signaling is also shown.
A major roadblock that prevents the systematic investigation of the contributions of RTKs and their signaling network to epilepsy is the slow speed of experiments designed to test the chronic effects of target inhibition in animal models of epilepsy. We have recently reported a novel in vitro antiepileptic drug screening platform that integrated an organotypic hippocampal culture model of epilepsy with microfluidic-multiple electrode array (µflow-MEA) chips [42]. The platform was capable of monitoring chronic epileptiform activity in inhibitor-treated cultures. A pilot screen conducted with the µflow-MEA platform revealed antiepileptic effects of EGFR/ErbB-2 and cFMS inhibitors. However, the transferability of the platform to other laboratories was limited due to complex clean room microfabrication requirements of the µflow-MEA chips. Here, we report a simplified screening platform that is based on standard tissue culture plates and commercially available microwire electrodes. We used this platform to evaluate the effects of 45 inhibitors of RTKs and their signaling pathways (Table 1) on the epileptiform activity in organotypic hippocampal cultures. We then used the data on kinase inhibitor selectivity to determine the likely kinase targets of the efficacious inhibitors.

Results
We screened 45 kinase inhibitors using the drug application and recording protocol shown in Figure 2. Inhibitors were screened with four cultures per inhibitor, and results were compared to four vehicle-treated cultures generated from the same animal. Seizure durations per 1 h of recording, average event rates, and electrographic loads were combined for all recordings and compared using a Kolmogorov-Smirnov test. The results for Flt-3 inhibitor [43] and cFMS receptor tyrosine kinase inhibitor (GW 2580) [44] are shown in Figure 3 as examples. These two inhibitors were also screened in our previous work [42] with similar results: cFMS inhibitor significantly reduced epileptiform activity in organotypic hippocampal cultures, while Flt-3 inhibitor had no significant effect. These results serve to validate the microwire screening platform. Western blot was conducted to confirm the suppression of kinase activity by corresponding inhibitors. Figure 3E shows a significantly decreased expression level of phospho-M-CSFR (cFMS) between control and cultures treated with cFMS inhibitor (GW2580). There was no detectable expression of phospho-flt3 in hippocampal slice cultures, which could potentially explain the lack of efficacy of Flt-3 inhibition. However, as discussed below, the lack of efficacy of a single inhibitor does not rule out Flt-3 as a potential antiepileptic drug target.

Results
We screened 45 kinase inhibitors using the drug application and recording protocol shown in Figure 2. Inhibitors were screened with four cultures per inhibitor, and results were compared to four vehicle-treated cultures generated from the same animal. Seizure durations per 1 h of recording, average event rates, and electrographic loads were combined for all recordings and compared using a Kolmogorov-Smirnov test. The results for Flt-3 inhibitor [43] and cFMS receptor tyrosine kinase inhibitor (GW 2580) [44] are shown in Figure 3 as examples. These two inhibitors were also screened in our previous work [42] with similar results: cFMS inhibitor significantly reduced epileptiform activity in organotypic hippocampal cultures, while Flt-3 inhibitor had no significant effect. These results serve to validate the microwire screening platform. Western blot was conducted to confirm the suppression of kinase activity by corresponding inhibitors. Figure 3E shows a significantly decreased expression level of phospho-M-CSFR (cFMS) between control and cultures treated with cFMS inhibitor (GW2580). There was no detectable expression of phospho-flt3 in hippocampal slice cultures, which could potentially explain the lack of efficacy of Flt-3 inhibition. However, as discussed below, the lack of efficacy of a single inhibitor does not rule out Flt-3 as a potential antiepileptic drug target.  The results for all 45 inhibitors are shown in Table 1 and graphically represented in Figure 4. The Pearson correlation coefficient was R 2 = 0.369 with p < 0.001 for cumulative seizure time versus electrographic load, and R 2 = 0.752 with p < 0.001 for cumulative seizure time versus average event rate. These results indicate that while the three metrics used to evaluate the results of the screen are The results for all 45 inhibitors are shown in Table 1 and graphically represented in Figure 4. The Pearson correlation coefficient was R 2 = 0.369 with p < 0.001 for cumulative seizure time versus electrographic load, and R 2 = 0.752 with p < 0.001 for cumulative seizure time versus average event rate. These results indicate that while the three metrics used to evaluate the results of the screen are not completely independent of one another, cumulative seizure time and electrographic load are only weakly correlated. Thus, both should be evaluated to determine inhibitor effects.  Kinase inhibitors are known to not always be specific to their intended target [21][22][23]. The suppression of epileptiform activity may occur through the inhibition of the intended target or through the inhibition of off-target kinases. Inhibition profiles of many of the kinases used in our screen have been published [21] or are available through manufacturers. We used this information to determine which kinases are inhibited by the compounds that suppressed epileptiform activity, but not inhibited by the compounds that had no effect. We selected nine "positive hits" (significant inhibition of epileptiform activity) and nine "negative hits" (no significant inhibition of epileptiform activity) with known inhibition profiles, indicated in Table 1 by green and orange highlighting, respectively. We then compiled a list of 182 kinases that were inhibited at a level of 50% or higher by concentrations of inhibitors that were close to the concentrations we used in the screen. We then counted the positive and negative hits against a given kinase, and sorted results by the difference between positive and negative hits (counts) ( Figure 5; complete results in Table S1). Kinase inhibitors are known to not always be specific to their intended target [21][22][23]. The suppression of epileptiform activity may occur through the inhibition of the intended target or through the inhibition of off-target kinases. Inhibition profiles of many of the kinases used in our screen have been published [21] or are available through manufacturers. We used this information to determine which kinases are inhibited by the compounds that suppressed epileptiform activity, but not inhibited by the compounds that had no effect. We selected nine "positive hits" (significant inhibition of epileptiform activity) and nine "negative hits" (no significant inhibition of epileptiform activity) with known inhibition profiles, indicated in Table 1 by green and orange highlighting, respectively. We then compiled a list of 182 kinases that were inhibited at a level of 50% or higher by concentrations of inhibitors that were close to the concentrations we used in the screen. We then counted the positive and negative hits against a given kinase, and sorted results by the difference between positive and negative hits (counts) ( Figure 5; complete results in Table S1).

Discussion
Six kinases (LYN, YES1, LCK, BLK, CDK5, and FYN) had three more inhibitors that targeted them and suppressed epileptiform activity than inhibitors that targeted them and did not affect epileptiform activity. Five out of six of these kinases (LYN, YES1, LCK, BLK, and FYN) are members of the Src family of non-receptor tyrosine kinases. LYN, YES1, and FYN are expressed in the human brain [45], while LYN, BLK, and FYN are expressed in the mouse hippocampus [33]. Src family kinases play an important role in learning and memory, and are involved in epilepsy [46,47]. CDK5, or cyclin dependent kinase 5, is also highly expressed in the human brain and mouse hippocampus, and plays an important role in the regulation of cytoskeletal organization, endocytosis, and exocytosis-processes important in epileptogenesis [33,45]. Thus, our approach of using the inhibition profiles of the screened inhibitors resulted in the identification of kinases that may be suitable targets for the suppression of epileptic seizures.
On the other hand, the screen also identified several kinases that were highly targeted by both effective and ineffective inhibitors of epileptiform activity. One example is NTRK2 (four positive hits, three negative hits). NTRK2 is neurotrophic receptor tyrosine kinase 2, also known as TrkB-a receptor of the brain-derived neurotrophic factor (BDNF). BDNF-TrkB signaling plays an important role in a variety of neuronal functions, and is involved in epilepsy [14][15][16][17]. Our results show that the inhibition of NTRK2 does not always reduce epileptiform activity, and that results are inhibitorspecific. Another example is CSF1R, colony stimulating factor 1 receptor, which had five positive and four negative hits on our screen. This tyrosine receptor kinase is also known as cFMS and is highly expressed in the hippocampus ( Figure 1A). CSF1R is the target of GW2580, a cFMS inhibitor that significantly reduced seizure load in organotypic hippocampal cultures in our previous work [42] and in this screen (Table 1). Interestingly, cFMS inhibitor also significantly inhibits the activity of NTRK1 and NTRK2 [21]. However, several 'negative hits', or inhibitors that did not significantly reduce epileptiform activity, also inhibit activity of CSF1R, NTRK1, and NTRK2 (IRAK-1/4 inhibitor,

Discussion
Six kinases (LYN, YES1, LCK, BLK, CDK5, and FYN) had three more inhibitors that targeted them and suppressed epileptiform activity than inhibitors that targeted them and did not affect epileptiform activity. Five out of six of these kinases (LYN, YES1, LCK, BLK, and FYN) are members of the Src family of non-receptor tyrosine kinases. LYN, YES1, and FYN are expressed in the human brain [45], while LYN, BLK, and FYN are expressed in the mouse hippocampus [33]. Src family kinases play an important role in learning and memory, and are involved in epilepsy [46,47]. CDK5, or cyclin dependent kinase 5, is also highly expressed in the human brain and mouse hippocampus, and plays an important role in the regulation of cytoskeletal organization, endocytosis, and exocytosis-processes important in epileptogenesis [33,45]. Thus, our approach of using the inhibition profiles of the screened inhibitors resulted in the identification of kinases that may be suitable targets for the suppression of epileptic seizures.
On the other hand, the screen also identified several kinases that were highly targeted by both effective and ineffective inhibitors of epileptiform activity. One example is NTRK2 (four positive hits, three negative hits). NTRK2 is neurotrophic receptor tyrosine kinase 2, also known as TrkB-a receptor of the brain-derived neurotrophic factor (BDNF). BDNF-TrkB signaling plays an important role in a variety of neuronal functions, and is involved in epilepsy [14][15][16][17]. Our results show that the inhibition of NTRK2 does not always reduce epileptiform activity, and that results are inhibitor-specific. Another example is CSF1R, colony stimulating factor 1 receptor, which had five positive and four negative hits on our screen. This tyrosine receptor kinase is also known as cFMS and is highly expressed in the hippocampus ( Figure 1A). CSF1R is the target of GW2580, a cFMS inhibitor that significantly reduced seizure load in organotypic hippocampal cultures in our previous work [42] and in this screen (Table 1). Interestingly, cFMS inhibitor also significantly inhibits the activity of NTRK1 and NTRK2 [21].
However, several 'negative hits', or inhibitors that did not significantly reduce epileptiform activity, also inhibit activity of CSF1R, NTRK1, and NTRK2 (IRAK-1/4 inhibitor, VEGFR tyrosine kinase inhibitor IV, and GTP-14564). It is therefore unclear whether the inhibition of CSF1R and NTRK2 is necessary or sufficient to suppress seizures based on the results of this screen.
The approach of using published inhibition profiles to identify effective kinase targets has several limitations. Published inhibition profiles were obtained in cell-free assays whereas our screen was conducted in a tissue-based model. We selected those inhibitors that were used at somewhat higher concentrations in our screen compared to the kinase activity assays. Higher concentrations would account for a potentially reduced availability of the inhibitor within cells in our organotypic model compared to cell-free assays. The published profiles do not cover all kinases; thus, there is a possibility that seizure suppression may be occurring through the effects on unidentified components of the kinome.

Microwire-Integrated Culture Plate
PFA-coated tungsten wire (bare diameter = 50.8 µm, coated diameter = 101.6 µm, A-M systems Inc.) was sterilized by 70% ethanol, affixed to the substrate of a standard tissue culture 6-well plate by silicone adhesive (4300 RTV, Bluestar Silicones, Oslo, Norway), and cured at 65 • C overnight. In each culture well, a recording electrode was fixed to the substrate with the microwire tip placed in the center and a reference electrode (same type of microwire with 1.5 cm insulation layer removed at the tip) was placed on the side ( Figure 2B). The microwire-integrated culture plates were then coated with poly-d-lysine (PDL, Sigma) and incubated in a humidified atmosphere at 37 • C overnight. Plates were then washed in sterile distilled water, filled with NeurobasalA/B27 medium (Thermo Fisher Scientific, Waltham, MA, USA) and incubated for at least 3 h before the placement of hippocampal slices.

Organotypic Hippocampal Slice Cultures
The hippocampi of postnatal day 7-8 Sprague-Dawley rat pups (Charles River Laboratories, Wilmington, MA, USA) were removed and cut into 350 µm slices on a McIlwain tissue chopper (Mickle Laboratory Eng. Co., Surrey, UK) and placed onto the microwire recording electrodes (with the electrode tip underneath the CA1/CA3 region) of the microwire-integrated culture plate, one slice per well. Slice cultures were maintained in serum-free NeurobasalA/B27 medium containing 0.5 mM glutaMAX (Thermo Fisher Scientific) and 30 mg/L gentamicin (Thermo Fisher Scientific) and incubated at 37 • C in 5% CO 2 on a rocking platform [48]. The medium was changed twice a week. All animal use protocols were approved by the Institution Animal Care and Use Committee (IACUC) at Lehigh University (Lehigh ID 167, originally approved on 5/5/2015 and re-evaluated on a yearly basis) and were conducted in accordance with the United States Public Health Service Policy on Humane Care and Use of Laboratory Animals.

Electrophysiology and Data Analysis
Organotypic hippocampal cultures were maintained in microwire-integrated culture plates for two weeks. Plates were transferred to a humidified mini incubator (maintained at 37 • C and 5% CO 2 ) for one-hour electrical recordings every other day ( Figure 2C). Microwire electrodes were connected to a multiple-channel amplifier (RZ2, Tucker Davis Technologies) with high-impedance head stage (PZ2-64, Tucker Davis Technologies; band-pass 1 Hz-3 kHz, gain ×1000). The sampling rate was 6 kHz per channel. Signals were processed and analyzed with OpenEx (Tucker Davis Technologies) and MATLAB (MathWorks), respectively.
We extended an automated algorithm for the quantification of durations of electrographically recorded seizures [49] to also quantify the rate of paroxysmal events. Two sets of analysis were performed: (1) bin duration was set to 0.5 s to quantify the duration of seizures in the recorded time period [49], (2) bin duration was set to 0.1 s to identify paroxysmal event rates within seizures up to a maximum of 10 Hz (tonic phase of electrographic seizures) ( Figure 2D,E) [50][51][52][53]. Ictal events (electrographic seizures) were defined as groups of paroxysmal events of much larger amplitude than background multiple unit activity and lasting longer than 10 s, including discrete shorter paroxysmal events that occurred with an event frequency of at least 2 Hz for at least 10 s. We used three parameters to determine the effects of inhibitors: seizure duration (time seizing during the recorded time period), average paroxysmal event rate (called event rate for conciseness), and electrographic load. Average event rate referred to the average over-threshold bin number (0.1 s binning in 10 s time window) in 45-min recordings (first 15 min of the 1-h recording was discarded). Electrographic load was the integration of the square of voltage with time.

Determination of Maximum Non-Toxic Inhibitor Concentration
We used the results of a lactate dehydrogenase (LDH) assay of culture supernatant and an analysis of the morphology of organotypic hippocampal cultures to determine whether an inhibitor was toxic at the applied concentration. Analyses were performed as previously described [42,48]. Briefly, all inhibitors were dissolved in DMSO and applied to n = 3 organotypic hippocampal slice cultures per concentration at 3 days in vitro (DIV), and morphology analysis and LDH assay were performed at 7 DIV. If LDH concentration or morphology scores were significantly elevated relative to vehicle (0.1% DMSO)-treated cultures, inhibitor concentration was considered toxic. The experiment was then repeated with a lower concentration. Maximum non-toxic concentrations of each inhibitor were used in the screen (Table 1).

Drug Application
Cultures from the same animal were organized into 3 experimental groups to test 2 drugs with a vehicle-treated control (n = 4 cultures per condition). All inhibitors were dissolved in DMSO at maximum non-toxic concentration and applied to cultures starting at 3 DIV. Control cultures were treated with 0.1% DMSO as vehicle. Inhibitors and vehicle were re-applied with each culture medium change. Electrophysiological data were then analyzed to evaluate drug efficacy. Sources of the drugs are provided in Supplementary Table S2.

Western Blots and Analysis
cFMS inhibitor GW2580 and Flt-3 inhibitor were applied to the culture medium after 3 days in vitro. After 24 h, cultures were collected from the 6-well plates and lysed in lysis buffer: RIPA buffer, phosphatase and protease inhibitors (Thermo Scientific, Waltham, MA, USA). Protein concentrations were assessed by micro BCA protein assay kit from Thermo Scientific. Proteins were separated in 12% Tris-Glycine Mini Gels (Life Technologies) and then transferred onto a PVDF membrane. Running and transfer buffers were purchased from Boston BioProducts. Firstly, for GW2580-treated cultures and the vehicle-treated control group the primary antibody Phospho-M-CSF receptor (Tyr723) Rabbit mAb was used for staining, and after stripping, primary M-CSF receptor antibody was applied to the membrane. For Flt-3 inhibitor-treated cultures and vehicle-treated cultures, initially primary antibody Phospho-FLT3 (Tyr589/591) Rabbit mAb was applied, and after stripping, FLT3(8F2) Rabbit mAb was applied (all primary antibodies from Cell Signaling). All primary antibodies were used at 1:1000 dilution. For all conditions Peroxidase-conjugated AffiniPure Goat Anti-Rabbit IgG (H+L) was used as the secondary antibody (Jackson ImmunoResearch). Bands were visualized on CL-XPosure X-ray films (Thermo Scientific) using SuperSignal West Femto Maximum Sensitivity Substrate. Quantification was performed using ImageJ software.

Conclusions
We developed an in vitro platform for determining the effects of compounds on chronic, spontaneous seizure-like activity. This microwire-based platform does not require sophisticated fabrication facilities and is transferrable to other laboratories. We used it to screen 45 kinase inhibitors and found several compounds with previously unknown anti-seizure effects. The examination of their kinase inhibition profiles led to the identification of kinases that may be promising targets for the development of antiepileptic drugs.

Conflicts of Interest:
The authors declare no conflict of interest.