The Fibroblast Growth Factor 14·Voltage-gated Sodium Channel Complex Is a New Target of Glycogen Synthase Kinase 3 (GSK3)*

Background: Fibroblast growth factor 14 (FGF14) binds to and regulates the voltage-gated Na+ (Nav) channel. Results: Inhibition of glycogen synthase kinase 3 (GSK3) modifies FGF14/Nav channel interaction, with effects on Na+ currents and subcellular distribution of the FGF14·Nav channel complex. Conclusion: The FGF14·Nav channel complex is a new target of GSK3. Significance: We provide evidence for modulation of Nav channels by GSK3 through FGF14. The FGF14 protein controls biophysical properties and subcellular distribution of neuronal voltage-gated Na+ (Nav) channels through direct binding to the channel C terminus. To gain insights into the dynamic regulation of this protein/protein interaction complex, we employed the split luciferase complementation assay to screen a small molecule library of kinase inhibitors against the FGF14·Nav1.6 channel complex and identified inhibitors of GSK3 as hits. Through a combination of a luminescence-based counter-screening, co-immunoprecipitation, patch clamp electrophysiology, and quantitative confocal immunofluorescence, we demonstrate that inhibition of GSK3 reduces the assembly of the FGF14·Nav channel complex, modifies FGF14-dependent regulation of Na+ currents, and induces dissociation and subcellular redistribution of the native FGF14·Nav channel complex in hippocampal neurons. These results further emphasize the role of FGF14 as a critical component of the Nav channel macromolecular complex, providing evidence for a novel GSK3-dependent signaling pathway that might control excitability through specific protein/protein interactions.

The intracellular FGFs (FGF11-14, also known as fibroblast growth factor homologous factors) are relevant components of the macromolecular complex of Nav 3 channels (Nav1.1-1.9) (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). Among this group of molecules, FGF14 stands out for its potent, specific, and diverse modulatory actions on Nav channels in the CNS, features that have been attributed to its unique nonconserved N-terminal domain (4 -7, 14). A number of in vitro experiments and studies in heterologous cells have examined the relative binding affinities of FGF14 to various Nav isoforms and evaluated the functional effects of FGF14 expression on Na ϩ currents, especially the ones encoded by the neuronal Nav1.2 and Nav1.6 isoforms (6). It has been shown that through high affinity binding to the intracellular C terminus of the ␣ subunit of Nav channels (7), FGF14 potently modulates amplitude and voltage dependence of Na ϩ currents, producing functional outcomes on Na ϩ currents of magnitude and direction that depend upon the channel isoform and the cell background (4 -6).
As part of the macromolecular complex of proteins at the axonal initial segment (AIS), FGF14 also serves to cluster Nav channels at the AIS, preserving action potential firing (5). Evidence comes from studies in hippocampal neurons where expression of a dominant negative FGF14 mutant protein suppresses Na ϩ currents and reduces expression and clustering of Nav channels at the AIS, leading to impairment of neuronal excitability (5). Consistent with these findings are the animal studies in fgf14 Ϫ/Ϫ mice showing that genetic deletion of fgf14 leads to decreased excitability (12,13) and impaired synaptic transmission (15), resulting in complex motor and cognitive deficits (16,17). Finally, expression of the naturally occurring FGF14 F145S missense mutation in humans results in ataxia, severe mental retardation, and neurodegeneration (18,19), indicating an evolutionary conserved role of FGF14.
Altogether these studies provide compelling evidence for FGF14 as a complex multivalent molecule vital for brain function. However, the intracellular signaling mechanisms regulating the assembly and stability of this protein/protein interaction complex remain largely unknown. We posited that identifying cellular pathways upstream of the FGF14⅐Nav channel complex would provide new fundamental knowledge on the mechanisms decoding intracellular signaling into functional in DMSO at 6 mM (stock solution) were automatically dispensed into the 96-well plates at 30 M final concentration in the medium (final concentration of DMSO 0.5%) using a Tecan Freedom EVO 200 liquid handling system (Tecan US, Durham, NC). Treated cells were incubated for 1 h at 37°C, at which time the reaction substrate D-luciferin, dissolved in serum-free phenol red-free DMEM/F-12 medium (final concentration 0.75 mg/ml), was automatically dispensed into the 96-well plates with a Synergy TM H4 Multi-Mode Microplate Reader (BioTek, Winooski, VT). Following 15 min of incubation, luminescence values per well (relative luminescence units, integration time of 1 s) were acquired with the Synergy TM H4 Microplate Reader. The cells were maintained at 37°C throughout the screening. Luminescence values expressed as percent average signal intensity from two independent screenings were normalized to per plate control (CLuc-FGF14 ϩ CD4-Nav1.6-NLuc treated with 0.5% DMSO). Control experiments were performed as described above to test the influence of compounds on the luciferase enzymatic activity, except that cells were transfected with pGL3 (1 g per well) expressing full-length Photinus luciferase instead of the LCA constructs.
For all other LCA experiments, the final concentration of DMSO was maintained at 0.5% unless stated otherwise, and selective kinase inhibitors were added manually in the culture medium (1-50 M final concentration). Luminescence values per well (relative luminescence units, integration time 0.5 s) were acquired at 2-min intervals for 20 -30 min using the Synergy TM H4 Microplate Reader. Signal intensity for each well was calculated as an average value of peak luminescence derived from three adjacent time points and expressed as percent control. Dose-response curves and IC 50 values were obtained by fitting the data with a nonlinear regression (14).
Measurement of Cell Viability-The influence of compounds on cell viability was measured in transiently transfected HEK-293 cells expressing full-length Photinus luciferase using Cyto-Tox-Glo TM cytotoxicity assay (Promega, Madison, WI), according to the manufacturer's instructions.
siRNA Gene Silencing of GSK3-To selectively silence GSK3, siRNA against GSK3 ␣/␤ was obtained from Cell Signaling (siRNA 6301) along with an unrelated siRNA used as negative control (Ambion 16706). For knockdown experiments, HEK-293 cells were plated in 6-or 24-well plates (50% confluency) and incubated 1 day later with 10 l of siRNA or negative control siRNA (both from a 10 M stock) using RNAiMAX (Invitrogen) at a ratio of 10:5 (l/l) according to the manufacturer's instructions. Two days later, cells were replated in either 24-or 96-well plates and transfected with a mixture of cDNA plasmid and with a second pulse of siRNA using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Forty eight h later, LCA and/or Western blot analyses were performed to confirm reduction in the complementation of CLuc-FGF14 and CD4 Nav1.6-C-tail-NLuc and silencing of GSK3.
Immunoprecipitations-Immunoprecipitations from HEK-Nav1.2 cells were essentially as described previously (5). For immunoprecipitations from rat brain, crude membrane extract was prepared from adult rat brains and homogenized as described previously (22) in a 0.3 M sucrose, 10 mM sodium phosphate monobasic homogenization buffer, 1 mM EDTA, pH 7.4, containing the following protease inhibitors: leupeptin (1 g/ml), aprotinin (1 g/ml), pepstatin (1 g/ml), and PMSF (1 mM). The homogenate was then centrifuged for 10 min (3000 ϫ g) at 4°C to remove nuclear debris, and the supernatant was centrifuged for 90 min (45,000 ϫ g) at 4°C; the resulting pellet (crude membrane extract) was resolubilized in the homogenization buffer described above. Crude membrane extract protein concentrations were measured by BCA (Pierce). Cell membrane extract was solubilized at 1:10 (ϳ1 mg/ml) in a lysis buffer consisting of: 1% Triton X-100, 0.15 M NaCl, 1 mM EDTA, 10 mM sodium azide, 10 mM Tris-HCl, pH 8.0, 2 mM NaF, BSA (1 mg/ml), aprotinin (1.5 g/ml), antipain (10 g/ml), leupeptin (10 g/ml), benzamidine (0.1 mg/ml), and PMSF (1 mM) on a tube rotator for 30 min at 4°C (23). Next, the samples were centrifuged (16,000 ϫ g) for 30 min at 4°C to remove the insoluble fraction. Monoclonal antibody mouse anti-PanNav (mAb K58/35, Sigma) at 2.5 g/ml was added to the soluble fraction and incubated on a tube rotator at 4°C overnight. Then the antibody⅐Nav channel complex was immobilized with 250 l of slurry of protein-A/G magnetic beads (Pierce) prewashed three times in lysis buffer at 4°C for 2 h on a tube rotator. Following incubation, beads were washed six times in 1 ml of lysis buffer without BSA and eluted with 4% acetic acid by incubation at 40°C for 15 min . Immunoprecipitated protein concentrations  were measured with Nanodrop (Thermo Scientific, Wilmington, DE).
Electrophysiology Data Analysis-Current densities were obtained by dividing Na ϩ current (I Na ) amplitude by membrane capacitance. Current-voltage relationships were generated by plotting current density as a function of the holding potential. Conductance (G Na ) as calculated by Equation 1, where I Na is the current amplitude at voltage V m , and E rev is the Na ϩ reversal potential.
Steady-state activation curves were derived by plotting normalized G Na as a function of test potential and fitted using the Boltzmann Equation 2, G Na /G Na,max ϭ 1 ϩ e ͑͑Va Ϫ Em͒/k͒ (Eq. 2) where G Na,max is the maximum conductance; V a is the membrane potential of half-maximal activation; E m is the membrane voltage, and k is the slope factor. For steady-state inactivation, normalized current amplitude (I Na /I Na, max ) at the test potential was plotted as a function of prepulse potential (V m ) and fitted using the Boltzmann Equation 3, where V h is the potential of half-maximal inactivation, and k is the slope factor. Data analysis was performed using Clampfit9 software (Molecular Devices) and Origin 8.6 software (OriginLab, Northampton, MA).
Immunohistochemistry-Fgf14 Ϫ/Ϫ and fgf14 ϩ/ϩ mice (P30) and rat pups (P11) derived from timed pregnant Sprague-Dawley rat females (Charles River, Wilmington, MA) were sacrificed through isoflurane exposure followed by decapitation. Whole brains were removed, frozen in liquid nitrogen vapor, and stored at Ϫ80°C until sectioning. Coronal sections (5-10 m) were serially cut using a crytostat, mounted on Superfrost glass microscope slides (Fisher Scientific, Waltham, MA), and left to dry overnight at room temperature. Sections were then fixed in fresh acetone at Ϫ20°C for 8 -10 min and dried at room temperature for 30 min, as described previously (24). Samples were washed three times with TBS, blocked for 1 h in TBS containing 0.3% Triton X-100 and 3% normal goat serum, and incubated overnight at 4°C with the following primary antibodies: mouse anti-FGF14 (monoclonal 1:100; Sigma) and rabbit anti-PanNav (polyclonal 1:100; Sigma) dissolved in TBS containing 1% normal goat serum and 0.3% Triton X-100. Samples were then washed and incubated for 1 h with the following secondary antibodies at a 1:250 dilution as follows: Alexa 488conjugated goat anti-mouse, Alexa 568-conjugated goat antirabbit, and Alexa 647-conjugated goat anti-chicken (Molecular Probes, Eugene, OR). In the experiments involving mouse tissue, 4Ј,6Ј-diamino-2-phenylindole⅐2HCl nuclear staining was performed according to the manufacturer's instruction (Santa Cruz Biotechnology). Samples were then mounted on glass slides with Prolong Gold anti-fade reagent (Invitrogen). Confocal images were then acquired with a Zeiss LSM-510 Meta confocal microscope with a ϫ20 air objective (0.75 numerical aperture), a ϫ40 (0.95 correction) air objective, and a C-Apochromat ϫ63 (1.2-watt corr) water objective. Multitrack acquisition was performed with excitation lines at 488 nm for Alexa 488, 543 nm for Alexa 568, and 364 nm for DAPI. Respective emission filters were band pass 505-530 nm, band pass 560 -615 nm low pass 650, and 385-470 nm. The optical slices were 1 m (mice) and 2 m (rats); for rat brain Z-stacks were collected at z-steps of 1 m with a frame size of 1024 ϫ 1024, pixel time of 1.26 s (ϫ20 objective) or 1.60 s (ϫ40 objective), pixel size 0.22 ϫ 0.22 m, and an 8-frame Kallman-averaging. For mouse brain, Z-stacks were collected with ϫ63 water objective at z-steps of 0.5 m with a frame size of 512 ϫ 512, pixel time of 3.2 s, pixel size 0.40 ϫ 0.40 m (0.7ϫ zoom), 0.14 ϫ 0.14 m (2ϫ zoom), and a 4-frame Kallman-averaging.
Primary Hippocampal Cultures-Banker's style hippocampal neuron cultures (25) were prepared from embryonic day 18 (E18) rat embryos using previously described methods (5). Briefly, following trituration through a Pasteur pipette, neurons were plated at low density (1-5 ϫ 10 5 cells/dish) on poly-Llysine-coated coverslips in 60-mm culture dishes in minimum Eagle's medium supplemented with 10% horse serum. After 2-4 h, coverslips (containing neurons) were inverted and placed over a glial feeder layer in serum-free minimum Eagle's medium with 0.1% ovalbumin and 1 mM pyruvate (N2.1 media; Invitrogen) separated by ϳ1 mm wax dot spacers. To prevent the overgrowth of the glia, cultures were treated with cytosine arabinoside (5 M; Calbiochem) at day 3 in vitro (DIV).
Immunocytochemistry-Hippocampal neurons (DIV14) were treated with either kinase inhibitors or DMSO (0.5%) for 24 h and then fixed in fresh 4% paraformaldehyde and 4% sucrose in phosphate-buffered saline (PBS) for 15 min. Following permeabilization with 0.25% Triton X-100 and blocking with 10% BSA for 30 min at 37°C, neurons were incubated overnight at room temperature with the following primary antibodies: mouse anti-FGF14 (monoclonal 1:100; Sigma), rabbit anti-PanNav (1:100; Sigma), and chicken anti-MAP2 (polyclonal 1:25000; Covance, Princeton, NJ) diluted in PBS containing 3% BSA. In some experiments, a rabbit anti-MAP2 (polyclonal 1:10,000; Millipore) was used in combination with a chicken anti-␤-IV-spectrin (1:25,000, gift from Dr. M. Komada (Tokyo Institute of Technology, Tokyo, Japan)). Neurons were then washed three times in PBS and incubated for 45 min at 37°C with appropriate secondary antibodies as described for brain tissue staining. Coverslips were then washed six times with PBS and mounted on glass slides with Prolong Gold antifade reagent.
Image Analysis-Confocal images were acquired with a Zeiss LSM-510 Meta confocal microscope with a ϫ63 oil immersion objective (1.4 NA). Multitrack acquisition was done with excitation lines at 488 nm for Alexa 488, 543 nm for Alexa 568, and 633 nm for Alexa 647. Respective emission filters were band pass 505-530 nm, band pass 560 -615 nm, and low pass 650. Optical slices were 2 m for the first dataset and 0.8 m for the second dataset; Z-stacks were collected at z-steps of 1 m for the first dataset and 0.4 m for the second dataset with a frame size of 512 ϫ 512, pixel time of 2.51 s, pixel size 0.28 ϫ 0.28 m, and a four-frame Kallman-averaging. All acquisition parameters, including photomultiplier gain and offset, were kept constant throughout each set of experiments. Acquired Z-stacks were sum-projected using MetaMorph 7.1.3.0 (Molecular Devices, Sunnyvale, CA), and pixel intensity values from the resulting stacked TIFF images were analyzed with ImageJ (National Institutes of Health). A region of interest (ROI) corresponding to a line of 6 pixels (1.68 m) in width and 72 pixels (20 m, first set) or 150 pixels (42 m, second set) in length was highlighted on an overlay image of MAP2, PanNav channel, and FGF14 staining along the MAP2-negative and FGF14-and PanNav channel-positive processes. The "starting point" for the ROI was the point of reduced MAP2 staining intensity that corresponds to the AIS as described previously (4 -7). A similar ROI was created for the dendrites on the Alexa 647 images corresponding to the MAP2 staining. The dendrites were defined as MAP2 positive processes. For each analyzed cell, the ROI for both the AIS and dendrites was then transferred to the Alexa 488 and Alexa 568 images and used for quantification of the FGF14 and the PanNav channel fluorescence intensity values, respectively. Total fluorescence intensity of immunolabeled PanNav channels and FGF14 at the AIS and in the dendrites was measured from cells in each set of experiments. The axonal-dendritic ratio was defined as the ratio of total fluorescence intensity in the AIS divided by the total fluorescence intensity in the dendrite within the ROI of the same size and was used as an index of polarized distribution of FGF14 and the Nav channel. The red/green ratio was defined as the total fluorescence intensity of the red channel divided by the total fluorescence intensity of the green channel in either AIS or dendrites and represented the level of co-localization/ association of the Nav channel with FGF14. The intensity of off-cell background fluorescence was determined by measuring an area free of cells but close to the ROI and was numerically subtracted at the end of the analysis prior to fluorescence quantification. To combine data from independent sets of experiments, total fluorescence intensity values were normalized to the mean value of the control (DMSO-treated neurons) of that set. Intensity profiles of individual traces for each experimental group were generated by plotting pixel intensity values along the distance. Data were tabulated and analyzed with Excel, Origin 8.6, and SigmaStat (Jendel Corp., San Rafael, CA). Linear filters available in Adobe Photoshop were applied to images for illustration purpose only.
Statistical Analysis-Results were expressed as means Ϯ S.E. The statistical significance of observed differences among groups was determined by Student's t test and one-way ANOVA or the corresponding nonparametric tests, Mann-Whitney rank sum, and Kruskal-Wallis tests, respectively, based on the distribution of the samples underlying the populations. A p Ͻ 0.05 was regarded as statistically significant. Bonferroni or Dunnett's tests were used for ANOVA post hoc analysis.

Kinase Inhibitor Screening of the FGF14⅐Nav Channel
Complex-In previous studies, we introduced the LCA (26) to detect the assembly of FGF14 with the C-terminal tail of the Nav1.6 channel in live cells (14). In this assay, the C-terminal and the N-terminal fragments of the P. pyralis luciferase are fused, respectively, to FGF14, CLuc-FGF14, and a chimera expressing CD4 fused to the Nav1.6 C-tail, CD4-Nav1.6-NLuc (Fig. 1A). Upon transient co-expression of the two constructs in HEK-293 cells, the FGF14⅐Nav1.6 C-tail complex formation is detected upon the addition of the substrate D-luciferin ( Fig. 1, B and C). To determine the applicability of LCA for pharmacological screenings, we tested its sensitivity to dimethyl sulfoxide (DMSO) commonly used as a vehicle for small molecule inhibitors. We found that neither the CLuc-FGF14⅐CD4-Nav1.6-NLuc complex formation (Fig. 1, C and D, n ϭ 3, p ϭ 0.22, one-way ANOVA) nor the relative protein expression levels of the two constructs are significantly affected by DMSO (Fig. 1E, n ϭ 3, p ϭ 0.70 for Cluc-FGF14, p ϭ 0.89 for CD4-Nav1.6-NLuc, one-way ANOVA). These features combined with the assay sensitivity, reliability, and favorable signal-to-noise ratio characterized in previous studies (14) prompted us to conclude that LCA would be suitable for small molecule screenings.
To gain insights into how the FGF14⅐Nav channel complex may be regulated by intracellular signaling cascades, we screened a small molecule library of 385 kinase inhibitors against the CLuc-FGF14⅐CD4-Nav1.6-NLuc complex, seeking pathways that could ultimately be relevant for neuronal excit-ability. Using the protocol illustrated in Fig. 1F, compounds were screened twice at a nominal test concentration of 30 M (delivered in DMSO, 0.5% final concentration). Among the potential hits, our screen identified seven structurally heterogeneous compounds targeting GSK3, producing a significant reduction in the LCA signal compared with control (change in signal Ͼ2 S.D. of the mean in control; Fig. 1G, orange dots). These compounds, including indirubin-3Ј-monoxime, kenpaullone, and the GSK3 inhibitors XIII, XII, X, IX, and VIII ( Fig.  2A), were then run through a counter-screening to account for potential interference of test compounds with the luciferase enzymatic activity (27,28), and three were declared hits (Fig.  2B, black dots). Studies have shown that some GSK3 inhibitors can also block the GSK3 closely related cyclin-dependent kinases (29). However, neither roscovitine nor roscovitine S-isomer, two pan-cyclin-dependent kinase inhibitors included in our library, were effective against the CLuc-FGF14⅐CD4-Nav1.6-NLuc channel complex (79 Ϯ 23% S.D., n ϭ 4, and 98 Ϯ 23% S.D., n ϭ 2, respectively), providing further evidence for a GSK3-specific effect on our target. Identification of GSK3 inhibitors as hits was of particular significance due to the documented role of GSK3 in regulating neuronal polarity (30), synaptic plasticity (31), and ion channels (20) and considering its relevance as a therapeutic target against brain disorders (32,33).
GSK3 expression was then reduced using validated siRNA (Fig. 2G). As illustrated in Fig. 2H, the assembly of the CLuc-FGF14⅐CD4-Nav1.6-NLuc complex upon GSK3 silencing was significantly reduced compared with siRNA negative control (p Ͻ 0.001, n ϭ 4, Student's t test) to a level comparable with GSK3 inhibitor XIII (Fig. 2D). Taken together, these data demonstrate that our strategy allowed the discovery of a signaling pathway identifying the FGF14⅐Nav channel complex as a new target of GSK3, a highly relevant multifaceted kinase with a central role in the normal and diseased brain (20,(33)(34)(35).
FGF14-dependent Regulation of Nav1.6 Currents Is Modified by GSK3 Inhibitors-Previous studies have shown that FGF14 regulates amplitude and voltage dependence of Nav-encoded Na ϩ currents in cell lines (4,6). To determine whether pharmacological inhibition of GSK3 affected FGF14-dependent modulation of Na ϩ currents, HEK-Nav1.6 cells were transiently transfected with Gfp or Fgf14-Gfp and treated with either DMSO (0.15% final concentration, control group) or GSK3 inhibitor XIII (30 M, final concentration) 30 -60 min prior to the experiments. In agreement with previous studies, we found that HEK-Nav1.6 cells expressing FGF14-GFP show significantly lower Na ϩ current (I Na ) amplitudes than cells expressing GFP (Ϫ18.7 Ϯ 2.4 pA/pF, n ϭ 13, for FGF14-GFPexpressing cells, Ϫ102.3 Ϯ 15.6 pA/pF, n ϭ 10, for GFP-expressing cells, F(2,31) ϭ 37.36, p Ͻ 0.0001, one-way ANOVA, post hoc Bonferroni, see Fig. 3, A-C). Notably, GSK3 inhibitor   3C). As expected from previous studies (6), analysis of voltage dependence of I Na activation and steady-state inactivation revealed significant changes in the kinetics of Na ϩ currents in the cells expressing FGF14-GFP compared with control (Fig. 3, D and E, and Table 1). Interestingly, treatment with GSK3 inhibitor XIII rescues the depolarizing shift of the steady-state inactivation induced by FGF14-GFP expression (approximately ϩ10 mV) back to control levels (F(2,22) ϭ 6.68, p Ͻ 0.0054, one-way ANOVA, post hoc Dunnett's test; Fig. 3, D and E, and Table 1). Thus, pharmacological inhibition of GSK3 modifies regulation of Na ϩ current amplitude and kinetics by FGF14, providing functional relevance to our hypothesis. GSK3 Inhibition Reduces the Assembly of the FGF14⅐Nav1.2 Channel Complex and Controls FGF14-dependent Modulation of Nav1.2 Currents-To extend our studies to another neuronal Nav channel isoform regulated by FGF14 (6), HEK-Nav1.2 cells were transiently transfected with Fgf14 -6xmyc and treated with either GSK3 inhibitor XIII (30 M) or DMSO (0.5% final concentration) for 2 h prior to harvest, and immunoprecipitations were performed with anti-Myc-agarose beads. As illustrated in Fig. 4, the fraction of Nav1.2 co-immunoprecipitating with FGF14-6ϫmyc was significantly lower (n ϭ 4, p Ͻ 0.05, Student's t test) in cells treated with GSK3 inhibitor XIII compared with DMSO control. Importantly, the levels of FGF14-6ϫmyc and Nav1.2 expression in the cell lysate were similar in cells treated with GSK3 inhibitor XIII compared with DMSO (Fig. 4), indicating that pharmacological inhibition of GSK3 is not likely to affect stability and/or degradation of FGF14 and Nav1.2 channel, but rather they decrease binding between the two interacting proteins.
Similarly to the experiments illustrated in Fig. 3, whole-cell patch recordings were performed in HEK-Nav1.2 cells transiently transfected with Gfp or Fgf14-1b-Gfp and treated with either DMSO or GSK3 inhibitor XIII. In agreement with previous studies, we found that HEK-Nav1.2 cells expressing FGF14-GFP show significantly lower Na ϩ current amplitudes than cells expressing GFP (Ϫ28.83 Ϯ 7 pA/pF, n ϭ 11, for FGF14-GFP-expressing cells, Ϫ68.37 Ϯ 7.03 pA/pF, n ϭ 13, for GFP-  Table 1.  Fig. 5, A-C). Pretreatment with GSK3 inhibitor XIII antagonized this FGF14-mediated phenotype, rescuing the level of Na ϩ current amplitude back to control levels (Ϫ53.42 Ϯ 11.14 pA/pF, n ϭ 16, for FGF14-1b-GFP-expressing cells treated with GSK3 inhibitor XIII compared with Ϫ62.4 Ϯ 5.9 pA/pF, n ϭ 13, for GFP-expressing control cells; F(2,37) ϭ 3.3, p Ͼ 0.05, one-way ANOVA, post hoc Dunnett's, see Fig. 5C), although no significant changes of Nav1.2 kinetics were found across conditions (Fig. 5, D and E, and Table 2). The lack of effect of FGF14 on the kinetics of Nav1.2 is in disagreement with previous studies (6) and might be attributable to an effect of DMSO. 4 Thus, GSK3 exerts a modulatory action on the FGF14⅐Nav channel complex that depends upon the Nav channel isoform, revealing an additional layer of complexity underlying this newly discovered pathway. FGF14 and Nav Channels Form a Complex in Native Tissue-The role of neuronal FGF14 has been evaluated in neuronal cultures (4,5,15) and in animal models using genetic manipulations (12,13,15,16). Yet studies of the distribution of FGF14 expression in native tissue are still missing. To gain insights into the expression pattern and subcellular distribution of native FGF14 and Nav channels, rat brain slices were fixed and probed with a rabbit antibody against PanNav channels and a mouse monoclonal antibody against FGF14 (Fig. 6, A-O). In the hippocampus, confocal analysis revealed a strong co-localization of FGF14 and Nav channels at the AIS, especially in the CA3 region and in the dentate gyrus (Fig. 6, A-F and M-P); a similar co-localization pattern, albeit in fewer cells, was observed in the CA1 region and in the subiculum (Fig. 6, G-I and J-L, respectively).
We further confirmed that FGF14 and Nav channels form a complex in the mammalian brain using co-immunoprecipitation from whole brain tissue followed by Western blot analysis (Fig. 6, Q and R). In addition to Western blot analysis (Fig. 6Q), the anti-FGF14 antibody was validated with immunofluorescence labeling of brain tissue of fgf14 Ϫ/Ϫ mice (Fig. 7). Altogether, these results provide further evidence for FGF14 as an integral component of the Nav channel complex at the AIS (5) with distinct regional brain localization, supporting specialized functions for this protein complex.
GSK3 Inhibition Induces Subcellular Redistribution of the FGF14⅐Nav Complex in Neurons-To determine the impact of the GSK3 signaling on the stability and subcellular distribution of the native FGF14⅐Nav channel complex, primary hippocampal neurons (DIV14) were exposed to GSK3 inhibitor XIII (30 M) or vehicle (0.5% DMSO) for 24 h and processed for confocal immunofluorescence analysis. Neurons were fixed and triple-labeled with a mouse monoclonal antibody against FGF14, rabbit polyclonal antibody against PanNav, and a chicken antibody against the somato-dendritic marker MAP2 (Fig. 8, A-P). In controls, FGF14 and Nav channels are found highly expressed and clustered at the AIS (5) and expressed at low level in the dendrites (Fig. 8, A-D and I-L). Following the inhibitor treatment, a significant loss in the FGF14 and Nav channel axonal polarity and co-localization was observed as evident from individual confocal images (Fig. 8, E-H and M-P), fluorescence intensity profiles, and red/green signal ratios (Fig. 8, Q  and R and S, respectively). Quantitative immunofluorescence analysis of confocal stacks over a large population of cells from two independent cultures (n ϭ 58 and 63 analyzed cells per category, respectively) revealed a prominent subcellular re-distribution of the FGF14⅐Nav channel complex from the AIS to dendrites upon pharmacological inhibition of GSK3 (Fig. 9). We observed a significant loss of Nav channel immunofluorescence at the AIS (p Ͻ 0.001, Student's t test, Fig. 9A), although its content in the dendrites increased (p Ͻ 0.05, Mann-Whitney test, see Fig. 9B). Conversely, FGF14 expression increased significantly in the dendrites (p Ͻ 0.001, Student's t test, see Fig.  9B), while remaining overall unchanged in the AIS (p ϭ 0.522, Mann-Whitney test, see Fig. 9A). These changes had a significant impact on the degree of association of the Nav channel with the FGF14 protein, as indicated by a prominent reduction in the red/green ratio at both the AIS and in the dendrites upon exposure to GSK3 inhibitor XIII (Figs. 8S and 9C). Furthermore, a drastic decrease in the axonal-dendritic ratio of both the Nav channels and FGF14 was observed ( Fig. 9D; p Ͻ 0.001, Student's t test for both groups).
To confirm the observed phenotype, we extended our analysis to another GSK3 inhibitor, CHIR99021 (CT99021) (36,37). Treatment with 3 M CHIR99021 induced a phenotype very similar to the one observed with the GSK3 inhibitor XIII, except that the level of reduction of Nav channels at the AIS was not statistically significant (Fig. 9A, n ϭ 20, p ϭ 0.921, Mann-Whitney test), although the FGF14 content significantly increased (Fig. 9B, n ϭ 20, p Ͻ 0.01, Mann-Whitney test), probably reflecting slightly different pharmacological properties of the two compounds. Importantly, the resulting reduction in the axonal-dendritic polarity and in the level of co-localization of 4 M. Nenov, T. F. James, and F. Laezza, unpublished observations. FGF14 and Nav channels at the AIS and in dendrites was even more pronounced with CHIR99021 treatment (Fig. 9, C and D).
Notably, the subcellular distribution of ␤-IV-spectrin, a component of the AIS (38), was unaffected by the treatment with GSK3 inhibitor XIII (Fig. 10) or CHIR99021 (data not shown), indicating a specific effect of GSK3 inhibition on the FGF14⅐Nav channel complex.

DISCUSSION
In this study we applied a combination of bioluminescence screenings, co-immunoprecipitation, patch clamp electrophysiology, and quantitative immunofluorescence to identify a link between the FGF14⅐Nav channel complex and GSK3. This kinase has well documented roles in neuronal polarity (30) and synaptic plasticity (20,31), and it has been found dysfunctional in a plethora of brain disorders (33)(34)(35). Collectively, our results provide consistent evidence for a modulation of interaction between the Nav channel and its multivalent regulatory protein, FGF14, upon inhibition of GSK3.
Our discovery was driven by the results of a bioluminescence-based screening of kinase inhibitors, implementing the LCA (26). Building on previous studies (14), we designed a screening platform incorporating counter-screening and additional validation steps that improved assay specificity and performance. We envision our approach as a novel tool to survey protein/protein interactions in the broad context of signaling networks allowing the identification of new regulatory pathways relevant for ion channels.
One of the key advantages of screening a chemical library as opposed to genetically silencing kinases (e.g. with RNAi knockdown) is that it provides the opportunity to easily screen transiently transfected cells. The limitation of this approach arises from off-target activity and cell toxicity exhibited by many kinase inhibitors (39). Several lines of evidence argue against  Table 2. nonspecific effects of GSK3 inhibitors on the FGF14⅐Nav complex. First, all compounds declared as potential hits that had GSK3 as their primary target (seven compounds) were structurally heterogeneous, yet consistently reduced the assembly of the FGF14⅐Nav channel complex, indicating conserved activity across different families of compounds. Second, inhibitors of cyclin-dependent kinases, such as roscovitine, did not significantly affect the FGF14⅐Nav channel complex, ruling out the most common source of side effects of GSK3 inhibitors (29). Third, the reduction of the FGF14⅐Nav1.6 C-tail complex assembly was unlikely ascribed to protein degradation (Fig. 2G) or cell toxicity (ϳ151.6% cell viability compared with DMSO control; data not shown). Fourth, the reduction in the FGF14⅐Nav channel complex assembly was reproducible by silencing GSK3 with siRNA (Fig. 2H).
Our electrophysiological experiments highlight the functional relevance of our study and reveal a new layer of complexity underlying the mechanism of regulation of Nav channels by FGF14 through GSK3. Previous studies have reported that expression of FGF14 in heterologous cells leads to inhibition of peak amplitude of Na ϩ currents encoded by Nav1.6 and Nav1.2 channel isoforms (6) and that these changes are likely to result from modulation in the number of cell surface available channels (4,5). Importantly, although GSK3 inhibitors act synergistically with FGF14 further suppressing Nav1.6 current, they antagonize FGF14 modulation of Nav1.2 rescuing Na ϩ current amplitudes back to control levels. In contrast, GSK3 inhibitors antagonize the action of FGF14 on Nav1.6 kinetics (rescuing V1 ⁄ 2 of steady-state inactivation back to control levels, see Fig. 3D and Table 1), with no detectable effects on Nav1.2 biophysical properties (Fig. 5, D and E, and Table 2). Although future studies are needed to identify the molecular nature of these differences, these results clearly demonstrate Nav isoform-specific differences in the effect exerted by GSK3 on the FGF14⅐Nav channel complex. One can speculate that GSK3 inhibition might favor alternative binding modalities of FGF14 to the Nav1.6 channel (for example through tethering of the FGF14 N-terminal tail with other Nav1.6 intracellular domains) resulting in an almost complete block of Na ϩ currents and a rescue of the channel kinetics (steady-state inactivation). Alternatively, GSK3 inhibitors might simply decrease or destabilize the interaction of FGF14 with the Nav1.2 channel, opposing the FGF14mediated suppression of Nav1.2 currents. Given the specialized subcellular segregation (40) and the diverse properties of Nav1.2 and Nav1.6 (6), these intricate effects of GSK3 on the FGF14⅐Nav channel complex are expected to greatly impact neuronal excitability.
At present, it remains unknown whether FGF14 or Nav channels are the primary target of GSK3 or whether the action of GSK3 on the FGF14⅐Nav channel complex requires any intermediate proteins/signaling molecules. Inspection of the primary amino acid sequence reveals a number of predicted GSK3 phosphorylation motifs throughout FGF14 and in the intracellular domains of Nav channels, but further studies are required to validate any of these Ser/Thr sites as GSK3 substrate(s).
To further validate the significance of our newly identified signaling mechanism, we carried out a detailed analysis of con-  N, and P). Arrows in M-P point to areas of co-localization of FGF14 and Nav channels along the AIS of a CA3 pyramidal neuron. Overlay image of the green (FGF14) and red (PanNav channels) is shown (C, F, I, L, and P). The boxed region in P highlights a representative AIS with pixel intensity profiles of both channels shown on the side (Q). Scale bar, 10 m; g ϭ granule cell layer; h ϭ hilus; p ϭ stratum pyramidalis; r ϭ stratum radiatum. R, representative Western blot of brain homogenate obtained from fgf14 ϩ/ϩ and fgf14 Ϫ/Ϫ adult mice (equal amount of proteins per lane). S, co-immunoprecipitation of FGF14 and the Nav channel from native tissue. Crude membrane extract (Input; 1:25 dilute) from whole adult rat brain was used for immunoprecipitation with a rabbit PanNav channel antibody (IP:PanNav), IgG (rabbit anti-luciferase, control), or blank (beads only). Western blots of input and co-immunoprecipitated fraction using a mouse anti-PanNav (IB:PanNav) and a mouse anti-FGF14 (IB:FGF14) antibodies are shown. Representative blot is from n ϭ 4 independent experiments. focal images showing that pharmacological inhibition of GSK3 leads to decreased co-localization and subcellular redistribution of the native FGF14 and Nav channel complex, leading to reversal of its axonal-dendritic polarity in hippocampal neurons (Figs. 8 and 9). This effect was largely determined by a reduction of the Nav channel expression at the AIS with a con-comitant increase in the dendrites and an increase of FGF14 expression mostly observed in the dendrites. Slight differences in the effect of CHIR99021 versus GSK3 inhibitor XIII were observed especially for the level of FGF14 at the AIS, but they did not affect the net result on the protein complex association (red/green ratio) or reversal of axonal-dendritic polarity. Fluorescence intensity values along AIS and dendrites are expressed as % mean fluorescence intensity in the corresponding process. S illustrates the level of co-localization of the Nav ␣ subunit and FGF14 along the AIS and dendrites in neurons illustrated in I-L and M-P, expressed as the ratio between red and green fluorescence intensity (neuron exposed to DMSO, black; neuron exposed to GSK3 inhibitor XIII, gray). Note the marked decrease of the red/green ratio in both the AIS and dendrites of the neurons treated with the GSK3 inhibitor compared with the DMSO-treated control.
Diminished association and loss of axonal-dendritic polarity of FGF14 and Nav channels might be mechanistically linked. FGF14 is rich in potential motifs critical for cellular trafficking, and the Nav channel C-tail contains endocytic di-leucine motifs that are required for its axonal compartmentalization (41). Although speculative, uncoupling of FGF14 from the Nav channel complex upon GSK3 inhibition could (i) directly impair cellular trafficking of Nav channels, leading to their dendritic accumulation and loss in axonal polarity, and (ii) prevent the accessibility of the channel C-tail di-leucine motifs, blocking channel endocytosis in the dendrites and limiting its compartmentalization in the axon. Notably, under our experimen-tal conditions, the subcellular distribution of ␤-IV-spectrin, a resident component of the AIS (38), was unchanged upon GSK3 inhibition (Fig. 10), suggesting a specific action of GSK3 on the FGF14⅐Nav channel complex. However, future studies are required to examine our results in the light of the role of GSK3 in establishing and maintaining neuronal polarity (30,(42)(43)(44)(45).
Overall, these results further emphasize the significance of FGF14 and provide evidence for a novel GSK3-dependent signaling pathway that acts at the level of the Nav channel macromolecular complex through specific protein/protein interactions, advancing our understanding of the molecular Values are normalized to the average pixel intensity in the DMSO-treated control. A, treatment with GSK3 inhibitor XIII (n ϭ 59) causes a 23.8 Ϯ 3.5% reduction of Nav channel in the AIS compared with the control (n ϭ 63; ***, p Ͻ 0.001, Student's t test) without changing the level of FGF14 (p ϭ 0.522, Mann-Whitney test); CHIR99021 treatment results in a slight (7.73 Ϯ 4.6%) but not statistically significant decrease of the Nav channel content (left) in the AIS (n ϭ 20; p ϭ 0.921, Mann-Whitney test) and a 22.6 Ϯ 7.0% increase of FGF14 (right) (**, p Ͻ 0.01, Mann-Whitney test). B, GSK3 inhibition causes an increase of Nav channel (left) in the dendrites both in the case of GSK3 inhibitor XIII (16.5 Ϯ 5.8%; *, p Ͻ 0.05, Student's t test) and CHIR99021 (48.5 Ϯ 17.2%; ***, p Ͻ 0.001, Mann-Whitney test) treatment; FGF14 (right) is also significantly increased in the dendrites in the presence of both GSK3 inhibitor XIII (53.5 Ϯ 8.2%; ***, p Ͻ 0.001, Student's t test) and CHIR99021 (92.7 Ϯ 18.5%; ***, p Ͻ 0.001, Student's t test). C, data from A and B are expressed as red/green ratio (C) (for the AIS p Ͻ 0.001 both in the GSK3 inhibitor XIII group and the CHIR99021 group; for the dendrites p Ͻ 0.001 in both groups; Student's t test) and axo-dendritic ratio of FGF14 and the Nav channel (D). Ratios between the red (Nav channel) and green (FGF14) pixel intensity values of the confocal images reflect the relative abundance of these proteins in the AIS (left) and dendrites (right). The red/green ratio in the AIS is significantly reduced after incubation with GSK3 inhibitor XIII (0.95 Ϯ 0.03; ***, p Ͻ 0.001, Student's t test) and CHIR99021 (0.80 Ϯ 0.03; ***, p Ͻ 0.001, Student's t test) compared with the DMSO control (1.38 Ϯ 0.08). D, axo-dendritic ratio of FGF14 and the Nav channel is significantly decreased after incubation with GSK3 inhibitor XIII (2.57 Ϯ 0.14 compared with 4.40 Ϯ 0.30 in the control for Nav channels and 1.51 Ϯ 0.11 compared with 2.25 Ϯ 0.14 in the control for FGF14; ***, p Ͻ 0.001, Student's t test) and CHIR99021 (2.17 Ϯ 0.18 for Nav channels, 1.09 Ϯ 0.12 for FGF14; ***, p Ͻ 0.001, Student's t test) compared with the DMSO control. determinants of excitability in the normal and diseased brain.