Identification of Amino Acid Residues in Fibroblast Growth Factor 14 (FGF14) Required for Structure-Function Interactions with Voltage-gated Sodium Channel Nav1.6*

The voltage-gated Na+ (Nav) channel provides the basis for electrical excitability in the brain. This channel is regulated by a number of accessory proteins including fibroblast growth factor 14 (FGF14), a member of the intracellular FGF family. In addition to forming homodimers, FGF14 binds directly to the Nav1.6 channel C-tail, regulating channel gating and expression, properties that are required for intrinsic excitability in neurons. Seeking amino acid residues with unique roles at the protein-protein interaction interface (PPI) of FGF14·Nav1.6, we engineered model-guided mutations of FGF14 and validated their impact on the FGF14·Nav1.6 complex and the FGF14:FGF14 dimer formation using a luciferase assay. Divergence was found in the β-9 sheet of FGF14 where an alanine (Ala) mutation of Val-160 impaired binding to Nav1.6 but had no effect on FGF14:FGF14 dimer formation. Additional analysis revealed also a key role of residues Lys-74/Ile-76 at the N-terminal of FGF14 in the FGF14·Nav1.6 complex and FGF14:FGF14 dimer formation. Using whole-cell patch clamp electrophysiology, we demonstrated that either the FGF14V160A or the FGF14K74A/I76A mutation was sufficient to abolish the FGF14-dependent regulation of peak transient Na+ currents and the voltage-dependent activation and steady-state inactivation of Nav1.6; but only V160A with a concomitant alanine mutation at Tyr-158 could impede FGF14-dependent modulation of the channel fast inactivation. Intrinsic fluorescence spectroscopy of purified proteins confirmed a stronger binding reduction of FGF14V160A to the Nav1.6 C-tail compared with FGF14K74A/I76A. Altogether these studies indicate that the β-9 sheet and the N terminus of FGF14 are well positioned targets for drug development of PPI-based allosteric modulators of Nav channels.

Voltage-gated sodium (Nav) 2 channels are responsible for the initiation and propagation of the action potential in excitable cells. Nine isoforms of Nav channels (Nav1.1-Nav1.9) have been characterized functionally, and evidence for a tenth one (Na x ) has been provided (1-12). Nav channels are differentially expressed in organs, with Nav1.1, -1.2, -1.3, and -1.6 found primarily in the central and peripheral nervous systems, Nav1.4 in the adult skeletal muscle, Nav1.5 in cardiac muscle, and Nav1.7, -1.8, and -1.9 primarily in the peripheral nervous system (3,4,7,12,13). With such widespread expression, it is not surprising that numerous diseases have been ascribed to mutations of specific Nav channel isoforms (4,14). These include the Dravet syndrome and other types of epilepsy (15)(16)(17); pain-related syndromes, such as congenital insensitivity to pain (18,19), primary erythromelalgia (20), and paroxysmal extreme pain disorder (21,22); and cardiac arrhythmias with congenital long QT syndrome (LQTS) type 3 (23,24); and Brugada syndrome (25). Furthermore, SNPs and/or copy variants within Nav channel genes have been associated recently with autism (Nav1.2) (26). Nav channels blockers are currently used in combined therapy for bipolar disorder (27,28), depression (29,30), and schizophrenia (31), extending the role of Nav channels to virtually all brain disorders both neurological and psychiatric (14,26,32). Their centrality in the pathophysiology of so many disruptive diseases has made Nav channels key pharmacological target sites for antiepileptic, analgesic, antiarrhythmic, and psychiatric drugs (11,14,33,34). Unfortunately, current Nav channel blockers lack specificity, as they are directed against molecular domains conserved across all Nav isoforms. As such, therapies based on these medications can result in severe side effects, such as Stevens-Johnson syndrome, blood dyscrasias, and ataxia (35). Although some success has been achieved in developing more targeted therapeutics against Nav channels (36), there is still an unmet need to develop safe and potent Nav isoform-specific compounds.
The pore-forming ␣-subunit of Nav channels is composed of four homologous domains (I-IV), each consisting of six transmembrane ␣-helices (S1-S6) and an additional pore loop located between the S5 and S6 segments (3). The S5 and S6 transmembrane segments from each domain make up a central pore when assembled within a tetrameric configuration. Upon depolarization, the pore of the channel allows Na ϩ to rapidly enter the cell; subsequently the channel inactivates and then closes (2). When expressed in heterologous systems, the ␣-subunit is sufficient to recapitulate the basic functional properties of the channel, but the kinetics, voltage dependence, gating, cellular targeting, and trafficking of the channel are modified by the many accessory proteins that compose the channel macromolecular complex in native conditions. Besides the ␤-subunits, other relevant regulatory proteins have been identified. As yet, caveolin-3, CaMKII, connexin-43, telethonin, plakophilin, ankyrins, NEDD4, SAPs, syntrophin-dystrophin complex, and intracellular fibroblast growth factors (iFGFs) have been identified as Nav channel accessory proteins (11,24,(37)(38)(39)(40)(41). Some of these interactors have been confirmed as components of the proteome of native Nav1.2 in the brain (42). This rich macromolecular complex of native Nav channels offers a unique source of specific protein-protein interaction (PPI) sites that could serve as targets for drug development (43). This is a new direction in pharmacology that has paid off in cancer (44) and cardiovascular fields (45), but it is still at a nascent stage in neuroscience. In searching for PPI surfaces that could lead to the development of probes and drug-like molecules targeting Nav channels, we have identified FGF14, a member of the iFGF family, as a physiologically relevant accessory protein with implications for brain function and pathology in both animal models and humans (46 -48). FGF14 is an emerging diseaserelevant protein that was initially associated with neurological disorders such as ataxia (49) and, from more recent genomewide association studies (GWAS), as a potential risk factor for schizophrenia (47) and depression (46). Binding of FGF14 to Nav1.1, Nav1.2, and Nav1.6 exerts powerful effects on Na ϩ currents, producing phenotypes that are Nav isoform-dependent and distinct from those associated with other iFGFs (39 -41, 50 -53).
In addition to binding to Nav channels, iFGF can form dimers. Previous structural studies have proposed the existence of a common interface of all iFGF responsible for both iFGF⅐Nav complexes and iFGF:iFGF dimer formation (51,52). However, this hypothesis has never been tested systematically and might not hold for FGF14 given its unique primary sequence (at the N terminus) and modulation of Nav channels (54,55).
To search for differences at the FGF14⅐Nav1.6 complex and the FGF14:FGF14 dimer interface, we engineered modelguided mutations at the predicted FGF14 surface and applied the in-cell split-luciferase complementation assay (LCA) to evaluate the effects of these mutants on FGF14:FGF14 dimer formation and monomer binding to the Nav1.6 C-tail. Through patch clamp electrophysiology, we then showed that either a single alanine mutation at Val-160 or a double alanine mutation at Lys-74/Ile-76 is sufficient to abolish previously described functional modulations of Nav1.6 currents by FGF14 (54,56) but that full functional activity of FGF14 requires an intact Val-160. Complementary studies using intrinsic fluorescence spectroscopy of purified proteins confirmed that Val-160 and Lys-74/Ile-76 are required for FGF14 binding to the Nav1.6 C-tail but that a single alanine mutation at Val-160 is structurally more disruptive. Overall, Lys-74/Ile-76 and Val-160 might be part of druggable pockets to be utilized for drug development against Nav channels.

Experimental Procedures
Materials-D-Luciferin was purchased from Gold Biotechnology (St. Louis, MO) and prepared as a 30 mg/ml stock solution in PBS and stored in a Ϫ20°freezer. Anti-luciferase antibodies against the C termini (aa 251-550) and N termini (aa 1-107) were purchased from Santa Cruz Biotechnology (Dallas, TX) and NovusBio (Littleton, CO), respectively.
Molecular Modeling-The FGF14⅐Nav1.6 homology model was generated using the FGF13⅐Nav1.5⅐CaM ternary complex crystal structure (4DCK) as a template. The FGF14 (aa 71-218) and Nav1.6 (aa 1790 -1917) sequences were aligned with the crystal structure of the FGF13⅐Nav1.5 complex (PDB code, 4DCK), and a project Protein Data Bank file was created using DeepView/Swiss-PdbViewer (58). This file was submitted to the Swiss-Model Web server software (QMEAN is 0.808 of 1); subsequently the model was improved by energy minimization in the Chiron Web server (59) and validated by the MolProbity Web server (60) (MolProbity score is 1.56, 94th percentile). Similarly, the FGF14:FGF14 dimer model was generated using the FGF13:FGF13 dimer crystal structure (PBD code, 3HBW) as a template. The FGF14 target sequence (aa 71-218) and the FGF13 crystal structure were aligned using DeepView/Swiss-PDBViewer. The resulting PDB file (QMEAN is 0.652 of 1) was submitted to the Swiss-Model Web server to generate the FGF14 dimer homology model. The model obtained from the Swiss-Model Web server was further improved by energy minimization by the Chiron Web server (59) and subsequently validated by MolProbity (MolProbity score is 1.47, 96th percentile). FGF14 K74A/I76A ⅐Nav1.6, FGF14 V160A ⅐Nav1.6 C-tail, FGF14 K74A/I76A ⅐FGF14 K74A/I76A , and FGF14 Y158A/V160A ⅐ FGF14 Y158A/V160A in silico mutations in FGF14 were generated by the USCF Chimera molecular modeling suite (61), and the best rotamers were selected according to their side-chain tor-sion as well as probability values in the rotamer library. Subsequently, energy minimization of the models was done using the Chiron Web server (59).
Split-luciferase Complementation Assay-Twenty-four hours after transfection, cells were replated from the 24-well plate using a 0.04% trypsin:EDTA mixture dissolved in PBS. Suspended cells were centrifuged and seeded in white, clearbottom CELLSTAR Clear 96-well tissue culture plates (Greiner Bio-One) in 200 l of medium. The cells were incubated for 24 h, and then the growth medium was replaced with 100 l of serum-free, phenol red-free DMEM/F-12 medium (Invitrogen). The bioluminescence reaction was initiated by automatic injection of 100 l of D-luciferin substrate (1.5 mg/ml dissolved in PBS) using a Synergy TM H4 multi-mode microplate reader (BioTek, Winooski, VT). Luminescence readings were initiated after 3 s of mild plate shaking and performed at 2-min intervals for 20 min with integration times of 0.5 s. Cells were maintained at 37°C throughout the measurements. Detailed methods for LCA can be found in previous studies (57).
LCA Data Analysis-Relative luminescence units (in RLU) measured using a Synergy TM H4 multi-mode microplate reader were tabulated by well position and time point into Microsoft Excel. The signal intensity for each well was calculated as a mean value of peak luminescence measured at three adjacent time points; the calculated values were expressed as a percentage of mean signal intensity in the control samples from the same experimental plate. Statistical values were calculated as mean Ϯ S.E. unless otherwise specified. The statistical significance (p Ͻ 0.05) of different groups was determined by Student's t test, one-way ANOVA with post hoc Bonferroni's method, or Kruskal-Wallis one-way ANOVA on ranks with post hoc Dunn's method using SigmaStat (San Jose, CA) and GraphPad Prism (La Jolla, CA) software. Graphs were plotted in Origin 8.6 software (OriginLab Corp., Northampton, MA).
Protein Overexpression and Purification-cDNAs encoding FGF14 WT (NCBI reference sequence accession number NP_787125, aa 64 -252) or the C-terminal domain of Nav1.6 (NCBI reference sequence accession number NP_001171455, aa 1756 -1939) were subcloned into suitable pET bacterial expression vectors (pET28a-FGF14 and pET30a-Nav1.6) with a His 6 tag at the N-terminal site; these plasmids were a gift of Drs. Regina Goetz and Moosa Mohammadi (52). The mutation coding for FGF14 K74A/I76A and FGF14 V160A was generated by sitedirected mutagenesis and PCR using FGF14 as a template. Upon transformation with the corresponding cDNA clones, recombinant proteins FGF14 WT , FGF14 K74A/I76A , FGF14 V160A , and Nav1.6 C-tail were expressed in the bacterial strain Escherichia coli BL21(DE3) pLys (Invitrogen) after induction with 0.2 mM isopropyl thio-␤-D-galactopyranoside for 8 h at 15°C. After induction with isopropyl thio-␤-D-galactopyranoside, bacterial cells were harvested and lysed by lysozyme and sonication at 4°C in lysis/binding buffer containing 10 mM sodium phosphate (prepared from 1 M Na 2 HPO 4 and NaH 2 PO 4 ) ϩ 0.1% CHAPS, pH 7.0 (for FGF14 proteins), 25 mM HEPES ϩ 150 mM NaCl ϩ 10% glycerol (Nav1.6), pH 7.5, containing 0.1 mM PMSF. The respective proteins were centrifuged at 18,000 ϫ g for 30 min at 4°C. For purification of FGF14 WT , FGF14 K74A/I76A , and FGF14 V160A , the supernatant was applied to pre-equilibrated heparin, and the proteins were then eluted with0.2-2.0 M NaCl in the lysis/binding buffer. For purification of the Nav1.6 C-tail, the supernatant was applied first to a Ni 2ϩ -nitrilotriacetic acid column and eluted with 200 mM imidazole. The Nav1.6 C-tail was further purified using a HiTrap Q Sepharose Fast Flow column (GE Healthcare) with a buffer containing 50 mM Tris-HCl and eluted with 10 -500 mM NaCl at pH 7.5. Finally, all concentrated proteins were purified on an AKTA purifier using Superdex 200 Hiload 16 ϫ 60 columns (both products from GE Healthcare Bio-Sciences) and equilibrated in 50 mM Tris-HCl ϩ 150 mM NaCl, pH 7.5.
Intrinsic Fluorescence Spectroscopy-The intrinsic fluorescence emission spectra of protein samples were recorded on a Spex FluoroMax (Spex Industries, Edison, NJ) in 25 mM Tris-HCl ϩ 150 mM NaCl, pH 7.5. The spectra (300 -450 nm) were recorded at an excitation wavelength of 295 nm in the proper spectral mode of the instrument using excitation and emission band passes of 5 nm each. The binding affinity of FGF14 WT , FGF14 K74A/I76A , and FGF14 V160A with the Nav1.6-Ctail was FGF14 Structure-Function Interactions with Nav1.6 determined by titrating the protein solutions with standard 5.5 M concentration aliquots.
Electrophysiology Experiments and Data Analysis-HEK-Nav1.6 cells transfected with GFP, FGF14-GFP, FGF14 V160A -GFP, FGF14 Y158A/V160A , or FGF14 Y158N/V160N -GFP were plated at low density on glass coverslips for 3-4 h and subsequently transferred to the recording chamber. Recordings were performed at room temperature (20 -22°C) 24 h post-transfection using a MultiClamp 700B amplifier (Molecular Devices, Sunnyvale, CA). The composition of the recording solutions consisted of extracellular salts (mM) (140 NaCl, 3 KCl, 1 MgCl 2 , 1 CaCl 2 , 10 HEPES, and 10 glucose, pH 7.3) and intracellular salts (mM) (130 CH 3 O 3 SCs, 1 EGTA, 10 NaCl, and 10 HEPES, pH 7.3). Membrane capacitance and series resistance were estimated by the dial settings on the amplifier and compensated for electronically by 70 -80%. Data were acquired at 20 kHz and filtered at 5 kHz prior to digitization and storage. All experimental parameters were controlled by Clampex 9.2 software (Molecular Devices) and interfaced to the electrophysiological equipment using a Digidata 1200 analog-digital interface (Molecular Devices). Voltage-dependent inward currents for HEK-Nav1.6 cells were evoked by depolarizations to test potentials between Ϫ100 and ϩ60 mV from a holding potential of Ϫ70 mV followed by a voltage pre-step pulse of Ϫ120 mV (Nav1.6). Steady-state (fast) inactivation of Nav channels was measured with a paired-pulse protocol. From the holding potential, cells were stepped to varying test potentials between Ϫ120 mV (Nav1.6) and ϩ20 mV (pre-pulse) prior to a test pulse to Ϫ20 mV. 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 ) was calculated by the following equation, 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, 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, where V h is the potential of half-maximal inactivation, E m is the membrane voltage, and k is the slope factor.
Transient I Na inactivation decay was estimated with a standard exponential equation. The inactivation time constant () was fitted with the following equation, where A 1 and ƒ 1 are the amplitude and time constant, respectively. The variable C is a constant offset term along the y axis. The goodness of fit was determined by the correlation coefficient (R), and the cutoff of R was set at 0.85.
Data analysis was performed using Clampfit 9 software (Molecular Devices) and Origin 8.6 software (OriginLab Corp.). Results were expressed as mean Ϯ S.E. The statistical significance of observed differences among groups was determined by Student's t test or one-way ANOVA with post hoc Bonferroni or Dunnett test; p Ͻ 0.05 was regarded as statistically significant.

Results
Homology Model-based Characterization of Putative FGF14 PPI Surface Hot Spots-To compare putative amino acid residues at the FGF14 PPI interface of the FGF14⅐Nav1.6 complex and the FGF14:FGF14 dimer formation, homology models based on other iFGFs (either in their dimeric form or in complex with the Nav1.5 channel C-tail) were created (Fig. 1, A and B, and Table 1). Inspection of the FGF14⅐Nav1.6 homology model revealed that in FGF14, residues Lys-74/Ile-76 (located at the N-terminal), Leu-116/Arg-117 (located at ␤-5), Asn-157/ Tyr-158/Tyr-159/Val-160 (located at ␤-9), and Leu-202/Pro-205/Val-208 (located at ␤-12) were within a distance of Ͻ8 Å (63, 64) from the closest neighboring amino acid of the Nav channel, consistent with the putative hot spots (Fig. 1C). Although most of these residues seemed to exert a similar role in the FGF14:FGF14 dimer complex, Lys-204 and Val-160 appeared to be structurally divergent (Fig. 1C). Lys-204 interacted with the neighboring Pro-205 in FGF14 but had no close neighbors in Nav1.6, whereas Val-160 interacted with residue Ile-1886 in Nav1.6 but had no putative interactors in the FGF14:FGF14 dimer. Thus, homology modeling predicts some conserved residues at the FGF14 surface, but potential structural differences depend on the local microenvironment.
In-cell Validation of Hot Spots at the FGF14⅐Nav1.6 and FGF14:FGF14 Dimer Interface-To evaluate the role of these model-based predicted hot spots experimentally, we engineered double/quadruple mutations in the FGF14 protein and examined their impact on the FGF14⅐Nav1.6 complex and FGF14:FGF14 dimer formation using our previously validated incell LCA (51,55,57,62,65). The FGF14 mutations were essentially grouped by ␤-sheet and/or N terminal location as in previous studies on FGF13 (52) and engineered to carry FGF14 mutant proteins fused with either CLuc (fused to the 5Ј-terminal end of the cDNA of interest) or NLuc fragments (fused to the 3Ј-terminal end of the cDNA of interest) of the Photinus pyralis firefly enzyme, allowing for in-cell reconstitution of FGF14⅐Nav1.6 C-tail and/or FGF14⅐FGF14 protein pairs. Mutations of the FGF14 protein considered in this study were the following: FGF14 K74F/I76R , FGF14 L116K/R117F , FGF14 N157D/Y159H , FGF14 Y158N/V160N , and FGF14 L202R/K204M/P205S/V208S . Combinations of FGF14 wild type (FGF14 WT ) or/and FGF14 mutant constructs (tagged with either CLuc or NLuc fragments) were transiently co-expressed with either CD4-Nav1.6-C-tail-NLuc (a chimeric construct that allows the membrane presentation of the Nav1.6 C-tail) (  with the same corresponding FGF14 mutant proteins in HEK293 cells (Fig. 2); this latter set of experiments was designed to reconstitute hetero-and homodimer forms of each FGF14 mutant . Upon binding of the respective protein pairs, the enzymatic activity of the luciferase enzyme was reconstituted by complementation of the full enzyme, giving rise to a robust luminescence response in the presence of the D-luciferin substrate. Representative luminescence responses of the assembly of CLuc-FGF14⅐CD4-Nav1.6-NLuc, CLuc-FGF14 K74F/I76R ⅐ CD4-Nav1.6-NLuc, and CLuc-FGF14 Y158N/V160N ⅐CD4-Nav1.6-NLuc are shown in Fig. 2A. For each construct pair, the maximum luminescence response of the CLuc-FGF14 mutant ⅐ CD4-Nav1.6-NLuc complex was normalized to the CLuc-FGF14 WT ⅐CD4-Nav1.6-NLuc complex (Fig. 2B). One-way ANOVA with post hoc Dunnett's analysis over a large data set (n ϭ 6 -9 independent experiments; n ϭ 4 repetitions) revealed that the strength of interaction of all protein complexes carrying mutations within the FGF14 protein was significantly reduced (p Ͻ 0.001) compared with the CLuc-FGF14 WT ⅐ Nav1.6-NLuc complex (Fig. 2B). LCA studies for both the FGF14 mutant homo-and heterodimer complexes with repre-FIGURE 1. Homology model-based predicted hot spots at the PPI interface of FGF14⅐Nav1.6 complex and FGF14:FGF14 dimer formation. A, the FGF14⅐Nav1.6 complex homology model (zoom view) was generated by FGF13⅐Nav1.5 (PDB code, 4DCK) crystal structure as a template. The C-tail of the Nav1.6 channel and FGF14 are shown as tan and gray colors, respectively. The critical amino acids Lys-74/Ile-76 (yellow), Leu-116/Arg-117 (magenta), Asn-157/Tyr-158/Tyr-159/Val-160 (green), and Leu-202/Lys-204/Pro-205/Val-208 (blue) are located at the N terminus, ␤5, ␤-9, and ␤-12 strands of FGF14. Critical amino acids of the C-tail of Nav1.6 channel are shown in red color. B, the FGF14:FGF14 dimer homology model (zoom view) was generated by FGF13:FGF13 (PDB code, 3HBW) dimer crystal structure as a template. C, the distance (less than 8 Å) between each critical amino acid of FGF14 to the neighboring critical amino acid of FGF14 or to the neighboring critical amino acid of the C-tail of the Nav channels was determined using UCSF Chimera software from homology models of the FGF14:FGF14 dimer and the FGF14⅐Nav1.6 complex. sentative traces and cumulative normalized luminescence responses are shown in Fig. 2, C and D. Importantly, as summarized in Fig. 2E, FGF14 mutations within the N terminus, ␤-5, and ␤-12 led to a relative decrease in PPI binding compared with control (Fig. 2, B-D, yellow, pink, and blue). Mutations of Val-160 and Tyr-158 resulted in reduced binding to the Nav1.6 C-tail (Fig. 2B), but they either had no significant effect on the relative binding strength (Y158N/V160N heterodimer) or augmented (Y158N/V160N homodimer) the relative binding strength when examined in the context of the FGF14 dimer (Fig. 2D). Furthermore, mutations at the Lys-74/Ile-76 residues had the greatest effect on both the FGF14⅐Nav1.6 complex and the FGF14 dimer, likely because of a strong interaction with neighboring amino acids as predicted by our molecular modeling. Western blotting analysis of total cell lysates derived from cells transfected with each pair of plasmids confirmed that the protein production across the experimental groups was com-parable, confirming the validity of the LCA results (Fig. 3). Altogether, these data support our homology model predictions suggesting structural divergence at the FGF14 ␤-9 sheet with mutations of Tyr-158/Val-160 and structural conservation at the FGF14 N terminus with mutations of K74A/I76A having a significant role in both the FGF14⅐Nav1.6 complex and the FGF14:FGF14 dimer formation.   signal of the FGF14 mutant ⅐CD4-Nav1.6 complex (normalized to FGF14 WT ⅐CD4-Nav1.6 complex) are shown in Fig. 4, A and B, and Table 2. One-way ANOVA with Dunnett's post hoc analysis revealed that one single Ala mutation at Tyr-158 was not sufficient to disrupt binding (CLuc-FGF14 Y158A ⅐CD4-Nav1.6-NLuc; 95.80 Ϯ 5.246%, n ϭ 21, p Ͼ 0.05), but a single Val-160 to Ala disrupted the complex (CLuc-FGF14 V160A ⅐CD4-Nav1.6-NLuc; 67.11 Ϯ 3.701%, n ϭ 21, p Ͻ 0.001). However, the double mutant exhibited a much lower relative binding (CLuc-FGF14 Y158A/V160A ⅐CD4-Nav1.6-NLuc; 33.63 Ϯ 2.0%, n ϭ 6) when compared with the FGF14 WT ⅐Nav1.6 complex (p Ͻ 0.001) or with the single V160A mutation (p Ͻ 0.001). The expression of all single and double Ala mutant proteins was confirmed across all groups by Western blotting analysis (Fig. 5,  A and B). Corresponding homology models of the FGF14 WT ⅐Nav1.6 and FGF14 V160A ⅐Nav1.6 complexes were built (Fig. 4, C and D) to inspect the role of Val-160 at the corresponding PPI interfaces. In the FGF14 WT ⅐Nav1.6 complex, Val-160 interacts with Ile-1886 (distance 4.1 Å) of Nav1.6 through a hydrophobic interaction (Fig. 4C). In the FGF14 V160A ⅐Nav1.6 model, the V160A mutation (orange) of FGF14 was further (red, 5.3 Å) from the Ile-1886 of Nav1.6 ( Fig. 4D), suggesting fewer opportunities for interaction with Nav1.6 (66).
When examined in the context of the FGF14:FGF14 dimer (Fig. 4, E and F, 49.85 Ϯ 2.05%, n ϭ 12, p Ͻ 0.01). The expression of all single and double alanine mutant proteins was validated across all groups by Western blotting analysis (Fig. 5, C and D). These results indicate that Val-160 alone is not sufficient to mediate PPI in the FGF14⅐FGF14 complex but that Val-160 might work synergistically with Tyr-158 to stabilize it. Corresponding homology models of FGF14 WT :FGF14 WT and FGF14 Y158A/V160A : FGF14 Y158A/V160A dimers were built (Fig. 4, G and H). The model predicts that Tyr-158 and Val-160 are within protruding ␤-9 loops that connect the two monomers in the intertwined dimer (Fig. 4G). Notably, Tyr-158 in each monomer appears to interact with Val-208 of the neighboring monomer through hydrogen bonding, whereas Val-160 had no predicted interactions. Simultaneous replacement of Tyr-158 and Val-160 with a neutral alanine residue might disrupt hydrogen bonding, weakening the stability of the ␤-9 loop (Fig. 4H). Homology modeling predictions and LCA results together suggest that the PPI interface of the FGF14⅐Nav1.6 complex is controlled by Val-160 through hydrophobic interactions, whereas the FGF14:FGF14 dimer requires the synergistic action of Tyr-158 and Val-160 through hydrogen bonding.
Intrinsic Fluorescence Spectroscopy Confirms a Key Role for Val-160 in the FGF14⅐Nav1.6 Complex-Our molecular modeling, LCA, and patch clamp data indicate that the Val-160 residue in FGF14 plays a unique and crucial role in modulating FIGURE 7. Protein production quantification from Western blotting for FGF14 K74A , FGF14 I76A , and FGF14 K74A/I76A . A, Western blotting analysis of whole-cell extracts from cells transfected with the indicated CLuc-FGF14 and CD4-Nav1.6-NLuc constructs. B, summary graph of densitometry analysis of CLuc and NLuc band intensity ration of the respective protein products. C, Western blotting analysis of whole-cell extracts from cells transfected with the indicated CLuc-FGF14 and FGF14-NLuc constructs. D, summary graph of densitometry analysis as described in C. Membrane were probed with anti-luciferase antibodies that recognize either the CLuc or the NLuc fragments (ϳ46 and ϳ66/114 kDa, respectively). Immunodetection of calnexin was used as loading control. Nav1.6 currents. A single alanine switch at this site or a concomitant alanine mutation at Tyr-158 and Val-160 are the only changes that can fully restore Nav1.6 currents to the GFP control level. Lys-74 and Ile-76, on the other hand, might work more synergistically, and mutations at these sites cannot completely rescue changes in Nav1.6 currents mediated by FGF14.
To provide correlative binding studies to our functional data, we used tryptophan-based fluorescence spectroscopy to probe  energy transfer processes occurring in PPI. The tryptophan fluorescence spectra for individual FGF14 WT , FGF14 K74A/I76A , FGF14 V160A , and Nav1.6-C tail proteins exhibited a max at 332 nm. Combining the FGF14 WT and the Nav1.6 C-tail increased the fluorescence emission intensity by more than 2-fold without any shift in the max, indicating strong protein complex formation without a change in the local environment. Both FGF14 K74A/I76A and FGF14 V160A mutants disrupted the interaction with Nav1.6-C-tail, but FGF14 V160A appeared to be the most damaging (Fig. 12). All mutant proteins had hydrodynamic radii identical to FGF14, as observed during gel filtration (data not shown). As evident from the emission spectra profiles (Fig. 12), none of the mutations lead to any major conformational changes in the protein complex, indicating that reduction in fluorescence intensity arises from decreased binding affinity (67).

Discussion
Previous studies have proposed that all iFGF might utilize a common interface for PPI with specific Nav isoform C-tail or other iFGFs (iFGF:iFGF dimer complexes); a hypothesis not yet directly tested. Through molecular, cellular, functional, and structural studies focused on FGF14, a disease-associated protein (46,48,49,68) and potent regulator of Nav1.6 channels (54, 56), we identified a significant structure-function similarity and divergence between the PPI interface within the FGF14⅐Nav1.6 complex and the FGF14:FGF14 dimer.
To provide experimental evidence for our model studies, we designed single, double, and quadruple mutations at the in silico predicted hot spots and tested FGF14 mutant activities using LCA to reconstitute PPI complexes in live cells (51,57,62,65). LCA studies confirmed our in silico predictions showing that most mutations destabilized both the FGF14⅐Nav1.6 and the FGF14⅐FGF14 complex but that the FGF14 Y158N/V160N double mutant led to opposite phenotypes depending on the structural context. The FGF14 Y158N/V160N mutant increased the stability of the FGF14:FGF14 dimer, but it impaired FGF14⅐Nav1.6 complex formation. Other amino acid residues that deserved attention were Lys-74 and Ile-76 in the N terminus of the FGF14. Mutations at these residues strongly impaired the FGF14⅐Nav1.6 complex formation, but the effect was preserved in the FGF14 dimer complex, indicating a potential conserved role for the N terminus at the two PPI interfaces.
This result prompted us to examine the role of Tyr-158, Val-160, Lys-74, and Ile-76 in the FGF14⅐Nav1.6 and FGF14:FGF14 dimer complex using targeted alanine-scanning mutations in combination with LCA. When examined in the context of the FGF14⅐Nav1.6 complex, I76A and Y158A did not lead to any phenotypes (Figs. 4B and 6B), whereas K74A and/or V160A were sufficient to disrupt FGF14⅐Nav1.6 complex formation ( Fig. 4B and 6B). In the FGF14 WT ⅐Nav1.6 homology model, we observed that Lys-74 and Val-160 interacted with Glu-1884 and Ile-1886 of Nav1.6, respectively, through salt bridge and hydrophobic interactions. These findings corroborate the critical role of Lys-74 and Val-160 residues in holding PPI interfaces through salt bridge (69) and hydrophobic interactions, respectively (70). Replacement of Lys-74 and Val-160 with a smaller alanine residue might increase the distance between the two neighboring residues, resulting in a less favorable structural environment for PPI (66). Importantly, we showed that FGF14 Y158A or FGF14 V160A alone are not sufficient to disrupt the FGF14:FGF14 dimer formation. In the FGF14:FGF14 dimer homology model, Tyr-158 directly interacts with Val-208 of the neighboring monomer via hydrogen bonding (71). Replacing Tyr-158 with Ala is not sufficient to interfere structurally with the dimer, but if combined with V160A the stability of the ␤-9 strand might be weakened and monomer affinity reduced. Simultaneous mutations of Tyr-158 and Val-160 to alanine can work synergistically to disrupt both the FGF14⅐Nav1.6 complex FIGURE 9. The V160A mutation abolishes FGF14-dependent modulation of biophysical properties of Nav1.6 currents. A, voltage dependence of I Na activation is plotted as a function of the membrane potential (mV); data (GFP, FGF14-GFP, and FGF14 V160A -GFP) were fitted with the Boltzmann function as indicated under "Experimental Procedures." B, box plot summary of V1 ⁄2 for voltage-dependent activation (voltage at which 50% of the channels are opened) in the indicated experimental groups. C, steady-state inactivation was measured using a two-step protocol, and values were plotted as a function of the membrane potential (mV). Data (GFP, FGF14-GFP, FGF14 V160A -GFP) were fitted with the Boltzmann function as indicated under "Experimental Procedures." The shift of voltage-dependent activation and steady-state inactivation is shown in the two insets in A and C, respectively. D, box plot summary of V1 ⁄2 for voltage-dependent steady-state inactivation (voltage at which 50% of the channels are closed) in the indicated experimental groups. Data are mean Ϯ S.E.; *, p Ͻ 0.05. with an Asn in each FGF14 monomer increases FGF14:FGF14 dimer formation (Fig. 2D). Both Tyr and Asn are polar residues; however, Asn is smaller than Tyr. Replacing both the bulky FIGURE 10. Functional validation of Lys-74 and Ile-76 in modulating Nav1.6 currents. A, representative traces of voltage-gated Na ϩ currents (I Na ϩ ) recorded from HEK-Nav1.6 cells transiently expressing GFP (gray), FGF14-GFP (black), FGF14 K74A -GFP (orange), FGF14 I76A -GFP (orange), and FGF14 K74A/I76A -GFP (green) in response to voltage steps from Ϫ120 mV to ϩ60 mV from a holding potential of Ϫ70 mV (inset). B, current-voltage relationships of I Na ϩ from GFP (gray), FGF14-GFP (black), and FGF14 K74A/I76A -GFP (green). C, box plot represents peak current densities measured in individual HEK-Nav1.6 cells expressing GFP, FGF14, FGF14 K74A -GFP, FGF14 I76A -GFP, and FGF14 K74A/I76A . D, representative traces of experimental groups described in A in which Tau () of I Na ϩ was estimated from a one-term exponential fitting function (red dotted line). Values are plotted as a function of amplitude and time constant. E, summary box plot of Tau calculated at the peak current density (Ϫ10 mV) in the indicated experimental groups. Data are mean Ϯ S.E. *, p Ͻ 0.05.

TABLE 5 Nav1.6-mediated currents in the presence of FGF14 and Lys-74 and Ile-76 mutants
The number of independent experiments is shown in parentheses.

Condition
Peak density Activation K act Inactivation K inact pA/pF mV mV mV mV ms Tyr-158 and the Val-160 residues in the two FGF14 monomers with a smaller Asn residue might facilitate interactions and increase the stability of the mutant homodimer (FGF14 Y158N/V160N : FGF14 Y158N/V160N ). These predictions and results are in agreement with previous in silico and LCA studies from our group (51). At the N terminus of FGF14, we found that Lys-74 directly interacts with Tyr-159 through a strong salt bridge and that replacing Lys-74 with an alanine disrupts this interaction, impairing the FGF14:FGF14 dimer formation. Lys-74 and Ile-76 acted synergistically in the FGF14⅐Nav1.6 complex formation but not in the FGF14 dimer (Fig. 6, A, B, E, and F), further supporting structural divergence at the two PPI interfaces.
Our molecular modeling and LCA studies were corroborated by whole-cell patch clamp electrophysiology. We investigated whether mutations at Lys-74 and Val-160 had any functional impact on the well described effect of FGF14 on Nav1.6-mediated currents. In agreement with previous studies, we found that FGF14 suppresses transient peak I Na ϩ density and affects voltage-dependent activation and steady-state inactivation compared with control. The single FGF14 V160A completely rescued peak current density to the control (GFP) (Fig. 8 and Table  4), whereas FGF14 K74A was only partially effective ( Fig. 10 and Table 5). Ile-76 worked synergistically with Lys-74 in that the FGF14 K74A/I76A mutant fully rescued Nav1.6-mediated currents to the control (GFP), supporting the LCA results (Fig. 6B). A more thorough analysis of Nav1.6 currents revealed a previously unreported effect of FGF14 on fast inactivation. This phenotype persisted with expression of FGF14 V160A and required the double mutations at Val-160 and Tyr-158 (FGF14 Y158A/V160A and FGF14 Y158N/V160N ) to be abolished. Notably, single lys-74 or Ile-76 or double Lys-74/Ile-76 mutations were unable to abolish the changes in Tau (Fig 10, D and E). Intrinsic fluorescence spectroscopy (72) based on purified proteins confirmed this model, indicating that the single V160A mutation was more disruptive than K74A/I76A in impairing the FGF14⅐ Nav1.6 complex formation.
Collectively, our studies demonstrate that amino acid residues located at the N terminus and at the ␤-9 of FGF14 are crucial for the FGF14⅐Nav1.6 complex and the FGF14:FGF14 dimer formation. Yet, the Val-160 residue is a point of divergence between the two complexes and is required for full FGF14 functional activity toward Nav1.6 channels. Although Val-160 is conserved in other iFGFs, its role varies depending on the structural environment provided by the specific iFGF and Nav channel isoforms. In the FGF13⅐Nav1.5 complex, for instance, Tyr-98 (the FGF13 residue corresponding to FGF14 Tyr-158) appears to have a more prominent role in the PPI complex formation, suggesting high precision and fidelity at each iFGF⅐Nav channel complex interface (69). Chemical probes that could leverage these unique structure-function features might provide an unprecedented opportunity for targeted interventions against excitability-driven brain and heart pathologies. FIGURE 11. Role of Lys-74 and I76 in modulating biophysical properties of Nav1.6 currents. A, voltage dependence of I Na activation is plotted as a function of the membrane potential (mV). Data (GFP, FGF14-GFP, and FGF14 K74A/I76A -GFP) were fitted with the Boltzmann function as indicated in under "Experimental Procedures." B, box plot summary of V1 ⁄2 for voltage-dependent activation (voltage at which 50% of the channels are opened) in the indicated experimental groups. C, steady-state inactivation was measured using a twostep protocol, and values were plotted as a function of the membrane potential (mV). Data (GFP, FGF14-GFP, and FGF14 K74A/I76A -GFP) were fitted with the Boltzmann function as indicated under "Experimental Procedures." The shift of voltage-dependent activation and steady-state inactivation is shown in the two insets in A and C, respectively. D, box plot summary of V1 ⁄2 for voltage-dependent steady-state inactivation (voltage at which 50% of the channels are closed) in the indicated experimental groups. Data are mean Ϯ S.E.; *, p Ͻ 0.05. FIGURE 12. Intrinsic fluorescence emission spectra reveal reduced assembly of FGF14 K74A/I76A and FGF14 V160A to Nav1.6 C-tail. Shown are the fluorescence spectra of the indicated purified proteins alone or combined; blank, Nav1.6, FGF14 WT , FGF14 K74A/I76A , FGF14 V160A , FGF14 WT ⅐Nav1.6, FGF14 K74A/I76A ⅐ Nav1.6, and FGF14 V160A ⅐Nav1.6 are shown as pink, red, blue, yellow, gray, black, green broken line, and orange broken line, respectively.
Author Contributions-S. R. A., A. K. S., and F. L. all wrote the manuscript. S. R. A. designed and performed the split-luciferase complementation assay, Western blotting analysis, and patch clamp electrophysiological experiments, analyzed the data, and built the homology models. A. K. S. purified the proteins and performed intrinsic fluorescence experiments. F. L. designed and supervised the work and the analysis and interpretation of the data. All authors read and approved the final version of the manuscript.