Modulation of the Voltage-gated Potassium Channel (Kv4.3) and the Auxiliary Protein (KChIP3) Interactions by the Current Activator NS5806*

Background: KChIP3 association with Kv4 channels regulate the K+ current gating. Results: NS5806 binds to KChIP3 and stabilizes the KChIP3-Kv4 complex. Conclusion: Ca2+ or NS5806 binding on KChIP3 decreases the dissociation rate of Kv4.3. Significance: The role of the hydrophobic cavity on KChIP3 in drug and protein association could lead to a better drug design for treatment of heart conditions. KChIP3 (potassium channel interacting protein 3) is a calcium-binding protein that binds at the N terminus of the Kv4 voltage-gated potassium channel through interactions at two contact sites and has been shown to regulate potassium current gating kinetics as well as channel trafficking in cardiac and neuronal cells. Using fluorescence spectroscopy, isothermal calorimetry, and docking simulations we show that the novel potassium current activator, NS5806, binds at a hydrophobic site on the C terminus of KChIP3 in a calcium-dependent manner, with an equilibrium dissociation constant of 2–5 μm in the calcium-bound form. We further determined that the association between KChIP3 and the hydrophobic N terminus of Kv4.3 is calcium-dependent, with an equilibrium dissociation constant in the apo-state of 70 ± 3 μm and 2.7 ± 0.1 μm in the calcium-bound form. NS5806 increases the affinity between KChIP3 and the N terminus of Kv4.3 (Kd = 1.9 ± 0.1 μm) in the presence and absence of calcium. Mutation of Tyr-174 or Phe-218 on KChIP3 abolished the enhancement of Kv4.3 site 1 binding in the apo-state, highlighting the role of these residues in drug and K4.3 binding. Kinetic studies show that NS5806 decreases the rate of dissociation between KChIP3 and the N terminus of KV4.3. Overall, these studies support the idea that NS5806 directly interacts with KChIP3 and modulates the interactions between this calcium-binding protein and the T1 domain of the Kv4.3 channels through reorientation of helix 10 on KChIP3.

ulation of I TO currents is a characteristic trait in hypertrophied and failing hearts (16) as well as in Brugada syndrome (17). Normal functioning heart tissue displays an increasing I TO current density gradient from endocardium to epicardium, which has been shown to correlate with a transmural KChIP density (18). Studies using canine models of failing hearts have shown that heart failure results in a decrease of phase 1 repolarization due to a decrease in I TO current as well as a remodeling of the gating kinetics of this current (19). Although numerous drugs have been shown to down-regulate the amplitude or modify the gating kinetics of I TO current, drugs that directly up-regulate the I TO current amplitude have not been reported.
Recently, Calloe et al. (17) introduced a new compound named NS5806, which was shown to modulate the I TO current kinetics and reverse the I TO current decrease induced by heart failure in canine models (20). Interestingly, this I TO activator has also been shown to decrease I SA currents in cultured hippocampal neurons where its effects on gating kinetics are comparable to those observed in heart tissue (11). Moreover, the induced potassium current modulation by NS5806 has been shown to depend on the presence of Kv4 auxiliary KChIP protein (11,21).
In this study we investigate the mechanism of NS5806 I TO / I SA current regulation by determining the interaction between NS5806 and KChIP3 and how this interaction modulates the KChIP3-Kv4.3 association. We show that the I TO activator NS5806 binds to the Kv4 auxiliary protein KChIP3 at the C terminus near the calcium binding sites, EF-hand 3 and EFhand 4, with an affinity comparable to the previously determined EC 50 for current potentiation. We also characterize the association constants between KChIP3 and peptide fragments of the T1 domain of Kv4.3 to show that calcium binding induces an increase in the affinity between KChIP3 and the T1 domain of Kv4.3. In the presence of NS5806 the interaction between KChIP3 and the hydrophobic N terminus of Kv4.3 is enhanced, and the calcium dependence of this interaction is abolished. Furthermore, anisotropy data suggest that the complex KChIP3-T1 domain adopts an altered quaternary structure in the presence of NS5806. Kinetic data indicates that NS5806 decreases the rate of dissociation between KChIP3 and site 1 of Kv4.3, which results in an overall increase in affinity. These results support the idea that NS5806 binds at the interface between the hydrophobic N terminus of Kv4.3 and the hydrophobic cavity at the C terminus of KChIP3 and stabilizes the protein complex. These results are further supported by docking, molecular dynamic simulations, and site-directed mutagenesis. Overall, the studies shown here provide a roadmap for the elucidation of the precise mechanism by which Kv4 current kinetics is regulated by KChIPs and NS5806.
Cells containing KChIP3-(65-256)-CCPGCC and GST-Kv4.3-(1-152) constructs were disrupted by sonication (Fisher, Model 100). However, these constructs were recovered as insoluble inclusion bodies that were collected by centrifugation. Inclusion bodies were washed extensively with 50 mM Tris buffer, pH 8.0, 1 M NaCl, 0.5% Triton X-100, 0.1% NaN 3 ,1 mM DTT, and 5 mM EDTA followed by a final wash with 50 mM Tris buffer, pH 8.0, and 2.0 M NaCl. Washed inclusion bodies were solubilized in 100 mM Tris buffer, pH 12.5, 2.0 M urea, and 5 mM DTT. Refolding was achieved by fast dilution of solubilized inclusion bodies into a 10ϫ volume of 100 mM Tris buffer, pH 7.4, 250 mM sucrose, 30% glycerol, 20 mM lauryldimethylamine N-oxide, 1 mM 2-mercaptoethanol on an ice bath with constant stirring. Refolded protein was concentrated and dialyzed against 20 mM Tris buffer, pH 7.4. Labeling of KChIP3-(65-256)-CCPGCC with the biarsenical probe FlAsH-EDT 2 was conducted by adding 2ϫ stoichiometric excess probe under identical condition as reported before (22). Unbound probe was removed by size exclusion chromatography, and labeled protein was further purified through a nickel-nitrilotriacetic acid column followed by extensive dialysis in 20 mM MOPS, pH 7.4, 100 mM NaCl. The labeling efficiency and stock concentration was determined by comparing the protein concentration using the bicinchoninic colorimetric method (Thermo Scientific) to the protein concentration determined using the FlAsH and ReAsH absorbance in 0.1 M NaOH. Protein purity and proper folding before FlAsH labeling was assayed using SDS-PAGE electrophoresis, CD spectroscopy, and the Zn 2ϩ -and Ca 2ϩinduced oligomerization changes as previously reported.  (22). The concentration of Ca 2ϩ and Mg 2ϩ was kept at 1 mM and 5 mM, respectively, and 1 mM EDTA or EGTA was added to obtain the calcium-free protein (herein referred as apo-state or apo-form) or Mg 2ϩ -bound KChIP3 samples, respectively.
Steady State Fluorescence Measurements-Steady state emission spectra were recorded using a PC1-ChronosFD spectrofluorometer (ISS). Tryptophan emission spectra were measured using a 295-nm excitation, and titrations were carried out in a 0.1 ϫ 1-cm path length quartz cuvette with excitation along the 0.1-cm path. KChIP3-NS5806 dissociation constants were determined by non-linear fit of the change in integrated fluorescence (310 -400 nm) using a single binding site model, where ⌬F is the intensity change, K d is the dissociation constant, n is the number of binding sites, P t is the total protein concentration, L t is the total ligand concentration, and c is a proportionality constant. For 1,8-ANS displacement studies, the excitation wavelength was set at 350 nm along the 2-mm path of a 0.2 ϫ 1-cm cuvette, and the integrated emission intensity (410 -500 nm) of 1,8-ANS bound to KChIP3 was used to calculate the extent of displacement. The dissociation constant of NS5806 for KChIP3 was recovered using a single binding site displacement equation (24), where, ␣ and ␤ are proportionality constants, K i is the dissociation constant of 1,8-ANS (measured separately), I t is the total concentration of NS5806, and K d is its dissociation constant.
To investigate the association between KChIP3 constructs and Kv4.3-(2-22) site 1 or Kv4.3-(70 -90) site 2, fluorescently labeled peptides were purchased from ThinkPeptides (Sarasota, FL). The fraction of Kv4.3-(2-22)-Ahx-FITC bound to KChIP3 constructs was determined based on the increase in steady state anisotropy and corrected for the increase in total intensity upon complex formation using, where f B is the fractional concentration of peptide bound, r is the measured anisotropy, r F and r B are the anisotropy of the free and fully bound peptide, respectively (25). Due to the large change in fluorescence intensity of Kv4.3-(70 -90)-Ahx-dansyl peptide upon binding to KChIP3 constructs, the fraction bound was determined by the change in integrated fluorescence intensity (450 -650 nm) normalized by the maximum intensity change. The resulting titration curves were analyzed using a single site binding model, Equation 1.
For the determination of the calcium dissociation constants, a sample containing 1 M KChIP3-(65-256) carrying the biarsenical binding sequence (CCPGCC) at the C terminus and labeled with FlAsH was prepared in 20 mM MOPS, pH 7.4, 100 mM NaCl, 1 mM DTT, 1 mM EGTA, and 1 mM nitrilotriacetic acid. All solutions were previously decalsified by treatment with CHELEX resin (Bio-Rad). The amount of residual calcium was determined using Quin-2 and calculated to be ϳ200 nM. Calcium binding was monitored by the change in integrated fluorescence emission (510 -580 nm) upon calcium binding to KChIP3-(65-256)-FlAsH. Calcium was added from a buffered 50 mM CaCl 2 stock prepared from a 100 mM CaCl 2 standard (Fisher), and the free calcium concentration was determined by obtaining the saturation curves for the calcium indicators Quin-2 (K d ϭ 120 nM), Calcium Green-5N (K d ϭ 14 M), and CaGreen-2 (K d ϭ 550 nM). The determined free calcium concentrations were in agreement with the values calculated using the pCa calculator software (26). The macroscopic calcium binding constants in the presence of NS5806 were obtained under identical conditions. FlAsH-labeled KChIP3 was excited with a 470-nm laser diode, and the change in fluorescence intensity was used to probe the extent of calcium binding. The resulting titration curves were analyzed using a double-Hill equation, where f bound is the fraction of proteins bound to calcium calculated from the intensity change, f is the fraction of sites B with dissociation constant k B and showing a Hill coefficient n b , with remaining sites showing a dissociation constant of k A and Hill coefficient n a . Fluorescence Lifetime and Modulated Anisotropy Measurements-Tryptophan fluorescence decay lifetimes were measured using a ChronosFD spectrofluorometer (ISS, Champaign, IL) in the frequency domain mode. Tryptophan was excited with the frequency modulated light (280-nm diode), and emission was collected using a 305-nm long pass filter and 400-nm short pass filter (Andover). The modulated anisotropy of tryptophan was calculated from the amplitude ratio of the parallel and perpendicular components of the modulated emission. The fluorescence decay lifetime of 1,8-ANS was determined by excitation with a 370-nm output of a frequency-modulated laser diode, and the emission was collected through 400 -600-nm wide pass filters (Andover). Modulation-phase data were best fitted by a multiple-exponential decay model using GlobalsWE software, and the 2 parameter was used as criterion for goodness of fit (27).
Isothermal Titration Calorimetry (ITC) Measurements-Thermodynamic parameters for NS5806 binding to KChIP3-(65-256) and KChIP3-(161-256) were determined using a VP-ITC titration calorimeter (Microcal Inc. Northampton, MA). ITC buffer (20 mM MOPS, pH 7.4, 0.25 mM DTT, and 100 mM NaCl) was prepared using decalcified water filtered through Chelex-100 resin (Bio-Rad). The protein stock solution was dialyzed against ITC buffer overnight. To minimize artifacts from mismatched buffers, stock NS5806 solutions were prepared in the final ITC dialysate buffer. The cell sample and injection syringe were extensively cleaned with decalcified water and then with ITC buffer. The reaction cell was loaded with 10 M KChIP3 (65-256 or 161-256) solution, and the concentration of NS5806 in syringe was 473 M. 30 injections (9 l each) of NS5806 were titrated into protein solution with 2-min intervals between injections. Parallel experiments were carried out for titration of NS5806 into the dialysate buffer as a control for heats of dilution of ligand. The temperature was kept at 25°C, and stirring speed was at 307 rpm. ITC data were analyzed using Origin 7 ITC data analysis software (OriginLab Corp., Northampton, MA). The data obtained for NS5806 binding to both KChIP3 constructs were analyzed using a two binding site model (28).
Molecular Modeling-The NS5806 binding sites on KChIP3 surface were identified using AutoDock 4.2 software (29). The structure of Ca 2ϩ -bound KChIP3-(65-256) (PDB entry 2JUL, conformation # 1) and KChIP3-(161-256) (PDB entry 2E6W) were used as a rigid macromolecule, and docking grids were set to cover the entire protein surface (15,30). The most favorable docking site among 200 docking simulation was selected based on their energy rankings. The structure of the NS5806⅐KChIP3 complex was further refined by running 10 ns of molecular dynamic simulation using AMBER03 force fields on the YASARA interface at 298 K, with a dynamic time step of 1.25 fs and trajectories recorded every 25 ps for further analysis (31). The molecular model for the KChIP3-Kv4.3 T1 domain was constructed by combining the crystal structure of KChIP1-Kv4.3 T1 domain (PDB entry 2NZ0) (32) and the KChIP3 or the NS5806⅐KChIP3 complex before molecular dynamic simulation. The energy-minimized KChIP3-Kv4.3 T1 domain or NS5806⅐KChIP3-Kv4.3 T1 domain complexes were then subjected to a 10-ns molecular dynamics simulation. The resulting protein and drug-protein complexes were aligned using the MUSTANG algorithm on the YASARA interface.

NS5806 Binds at the C Terminus of KChIP3-The
Kv4-mediated current amplitude increase and slowing of inactivation associated with NS5806 have been reported to be enhanced or dependent on the presence of the auxiliary KChIP proteins (11,21). In this study we probe whether this dependence is due to direct interaction between NS5806 and the neuronal calcium sensor KChIP3. NS5806 interaction with KChIP3 was initially characterized by monitoring the steady state tryptophan fluorescence emission of KChIP3 upon the addition of NS5806. The single tryptophan residue (Trp-169) found in KChIP3-(65-256) sequence is located at the interface between EF-3 and EF-2, and thus its emission may probe NS5806 association to either N-or C-terminal domains. Upon the addition of a 2ϫ excess of NS5806, a significant decrease in tryptophan emission intensity was observed (Fig. 1a). The quenching effect was present in the apo-state as well as in the calcium-bound state of KChIP3. The absorption spectrum of NS5806 in 20 mM MOPS buffer shows an absorption peak at 340 nm with an extinction coefficient of 105 M Ϫ1 cm Ϫ1 . This absorption band overlaps with the Trp-169 emission spectrum, suggesting that resonance energy transfer between Trp-169 and NS5806 is responsible for the observed Trp emission quenching.
To determine the KChIP3 affinity for NS5806, the Trp emission intensity was plotted as a function of NS5806 concentration of KChIP3-(65-256) (Fig. 1b), and the dissociation constants were recovered using Equation 1 (see Table 3). The stoichiometry of the model was set to 1 as analysis of the tryptophan fluorescence by the continuous variation method yielded a stoichiometry of one to one (data not shown). The recovered dissociation constants show that in the Ca 2ϩ -bound form, KChIP3-(65-256) binds NS5806 with a dissociation constant of 4.4 Ϯ 0.3 M, whereas a separate experiment using the truncated C terminus domain, KChIP3-(161-256), shows roughly a 2-fold affinity increase to K d ϭ 2.7 Ϯ 0.3 M (data not shown). Upon Ca 2ϩ removal, the affinity of KChIP3-(65-256) for NS5806 decreases, K d ϭ 30 Ϯ 2 M, whereas the affinity of KChIP3-(161-256) is reduced to K d ϭ 26 Ϯ 4 M. Because both constructs show similar affinities for NS5806 in the calciumbound form and apo-state we propose that the C terminus of KChIP3 is the binding site for NS5806.
It is interesting to note that the decrease in fluorescence of apoKChIP3-(65-256) upon the addition of NS5806 is about twice as large as that observed for apoKChIP3-(161-256) (data not shown), and we hypothesize that this could be due to a structurally different conformation of the isolated C terminus in the apo-state. To investigate if this is the case, the quenching data were analyzed using a modified Stern-Volmer equation, which accounts for the presence of tryptophan residues that may not be quenched by NS5806 (Fig. 1d). The results show that both KChIP3-(65-256) and KChIP3-(161-256) in the apoform bind to NS5806 with the same affinity and that the observed differences in the extent of quenching are due to the fact that in the apo-state, KChIP3-(161-256) populates a second conformation (ϳ40%) that is not quenched by NS5806. This additional population may be present due to a lower stability of the truncated C terminus domain.
To probe if resonance energy transfer is responsible for the observed fluorescence quenching, we measured the tryptophan fluorescence lifetime of both KChIP3 constructs in the calciumbound form. The average lifetime of Trp-169 decreases from 3.3 ns to ϳ1.3 ns in Ca 2ϩ KChIP3-(65-256) and from 4.3 ns to 3.2 ns in Ca 2ϩ KChIP3-(161-256) as the concentration of NS5806 is increased (shown in Fig. 1c). The fluorescence lifetime decrease indicates that the quenching mechanism happens during the excited state of tryptophan and correlates well with a dynamic quenching process such as resonance energy transfer.
Furthermore, it is possible to use the fluorescence emission spectra of both KChIP3 constructs, the tryptophan fluorescence quantum yield, and the absorbance spectra of NS5806 while assuming a random NS5806-Trp-169 orientation to obtain the overlap function and calculate the Förster distance between the NS5806-Trp-169 pair for KChIP3-(65-256) and KChIP3-(161-256). The Förster distances were calculated to be 12 Å and 14 Å for the apo-form KChIP3-(65-256) and KChIP3-(161-256), respectively, whereas a Förster distance of 12 Å was recovered for the Ca 2ϩ -bound form of KChIP3-(65-256) and 13 Å for Ca 2ϩ KChIP3-(161-256) ( Table 1). Due to the location of Trp-169 within the protein matrix, a random orientation value for 2 is not appropriate. However, an upper ( 2 max ) and lower ( 2 min ) limit for the orientation factor can be obtained from measurement of the frequency-modulated anisotropy of Trp-169 for both KChIP3 constructs (data not shown). These values were used to calculate the upper and lower limits for NS5806-Trp-169 distances as shown in Table 1. The recovered NS5806-Trp-169 distances using fluorescence emission and fluorescence lifetime of Trp-169 range from 5 Å to 13 Å and from 8 Å to 20 Å, respectively. The larger values for the lifetime measurements support the presence of an additional static quenching process induced by binding of NS5806, which results in smaller distances being resolved in the steady state quenching measurements. Additional contribution due to quenching of residual tyrosine fluorescence cannot be ruled out. Overall, these results show that NS5806 binds at the C terminus of KChIP3 near EF-hands 3 and 4 and quenches the single tryptophan residue through resonance energy transfer. NS5806 also showed binding and quenching of tryptophan residues on a Kv4.3-(1-152)-GST fusion protein; however, the dissociation constant for binding to this construct was 11 Ϯ 2 M (Fig. 1, e and f). The low solubility of the full T1 domain prevented us from further characterization of this interaction and from unequivocally showing that NS5806 binds at the T1  Table 3.   Ca 2ϩ KChIP complex did not alter the 1,8-ANS lifetimes; however, a systematic decrease in the pre-exponential factor for the long lifetime (␣ 2 ) and an increase of the pre-exponential parameter for 1,8-ANS in solution (␣ 0 ) were observed, whereas the pre-exponential factor for 1,8-ANS bound at the site with 6-ns lifetimes (␣ 1 ) remained fairly constant (Fig. 2c). These results indicate that NS5806 binds to the hydrophobic site/cavity on KChIP3, which results in a displacement of 1,8-ANS bound at that site.
NS5806 Binding to KChIP3 Is Enthalpy-driven-The thermodynamic properties associated with the formation of NS5806⅐KChIP3 complexes were probed using isothermal calorimetry. Isotherms for the titration of NS5806 into Ca 2ϩ KChIP3-(65-256) and Ca 2ϩ KChIP3-(161-256) in the  presence of 5 mM Mg 2ϩ are shown in Fig. 3. The isotherms were analyzed using a two-site binding model (solid line). The binding of NS5806 to KChIP3 was characterized with two dissociation constants with the specific binding site associated with a K d of 0.8 M and a nonspecific binding site that showed a very weak affinity with K d ϳ 150 M (data not shown). This weak binding site was not observed in the Trp-169 quenching and 1,8 ANS displacement studies, probably due to an inefficient quenching or different binding site than that of 1,8-ANS. ITC thermodynamic parameters for NS5806 specifically binding to KChIP3 constructs are summarized in Table 2. The association of NS5806 to KChIP3-(65-256) and KChIP3-(161-256) exhibited an exothermic binding with ⌬H ϭ Ϫ4.4 kcal mol Ϫ1 and ⌬H ϭ Ϫ6.9 kcal mol Ϫ1 , respectively. Association of NS5806 to KChIP3-(65-256) showed different enthalpic and entropic contributions than KChIP3-(161-256), which may indicate that the structural rearrangements induced by NS5806 are not localized solely on the C terminus domain. An exothermic interaction indicates that the overall interaction may involve a structural reorganization of the protein-drug complex upon association with the NS5806.  Fig. 4a, with the recovered parameters listed on Table 3.
Binding of site 1 to Ca 2ϩ KChIP3-(65-256) was marked by an increase in anisotropy from r ϭ 0.072 to r ϭ 0.238 and a strong calcium dependence with a dissociation constant for Kv4.3 site 1 in the apo-form of K d ϭ 70 Ϯ 3 M and K d ϭ 2.7 Ϯ 0.1 M in the calcium-bound form. A stronger calcium dependence was observed for binding of site 2 to KChIP3-(65-256), with a dissociation constant of ϳ500 M in the apo-state and K d ϭ 10 Ϯ 1 M upon calcium binding. Similarly, binding of Site 1 to the KChIP3-(161-256) construct, which lacks the N terminus domain, showed an increase in anisotropy to r ϭ 0.230 in the presence of calcium as well as a strong calcium dependence with K d ϳ450 M in the apo-form to K d ϭ 24 Ϯ 1 M in the calcium-bound form. On the other hand, binding of site 2 to calcium-bound KChIP3-(161-256) was weak (K d ϳ390 M), indicating that binding of this peptide is specific to the N terminus domain as proposed previously (32,36). Furthermore, upon the addition of 150 M NS5806, the binding affinity of  Table 4.   Table  3). The NS5806-induced affinity increase is dose-dependent and was observed in the apo-and Ca 2ϩ -bound forms (Fig. 4b).
Interestingly, in the presence of Ca 2ϩ and site 1 peptide, the affinity for NS5806 decreased significantly, likely due to a restricted access to the hydrophobic cavity ( Fig. 4 and legend).
On the other hand, the addition of Ca 2ϩ to the apoKChIP3-NS5806-site 1 complex did not result in any change in anisotropy. In the apo-form, the affinity between site 2 and KChIP3 was not observed to be modulated by NS5806. These results support the idea that the KChIP3 Kv4.3 T1 domain interactions are regulated by calcium and that in the presence of NS5806, KChIP3 undergoes structural rearrangements that favor the association of the hydrophobic N terminus of Kv4.3.

NS5806 Modulates Kv4.3 Site 1 Binding to KChIP3 by
Decreasing the Rate of Dissociation-To better understand the mode of action of NS5806, we investigated its effect on the kinetics of binding between KChIP3 and site 1 of Kv4.3. Association of the fluorescently labeled peptide analogous to site 1 of Kv4.3 with KChIP3 in the apo-form or calcium-bound form resulted in an ϳ18% increase in total fluorescence intensity. The fluorescence intensity increase can be attributed to a decrease in solvent accessibility of the fluorophore upon binding. The change in fluorescence intensity was used to probe the binding kinetics of Kv4.3 site 1 to KChIP3-(65-256) and KChIP3-(161-256); a representative trace is shown on Fig. 4c. The kinetic traces were best analyzed using a double exponential decay, and the recovered parameters are listed in Table 4. The association rate for the fast and slow phase of site 1 binding to KChIP3-(65-256) were similar in the apo-and calciumbound state, with k on fast ϭ 17 Ϯ 5 mM Ϫ1 s Ϫ1 and k on slow ϭ 2.5 Ϯ 0.3 mM Ϫ1 s Ϫ1 in the apo-state and k on fast ϭ 28 Ϯ 8 mM Ϫ1 s Ϫ1 and k on slow ϭ 2.6 Ϯ 0.7 mM Ϫ1 s Ϫ1 in the calcium-bound form. However, the pre-exponential factors associated with the slow (␣ 2 ) phase showed an increase upon binding of calcium from 0.02 in the apostate to 0.13 in the calcium-bound state. The parameters most influenced by binding of calcium were the dissociations rates of both phases, decreasing from k off fast ϭ 1170 Ϯ 30 ϫ 10 Ϫ3 s Ϫ1 and k off slow ϭ 2.9 Ϯ 0.7 ϫ 10 Ϫ3 s Ϫ1 in the apo-form to k off fast ϭ 77 Ϯ 20 ϫ 10 Ϫ3 s Ϫ1 and k off slow ϭ 0.2 Ϯ 0.06 ϫ 10 Ϫ3 s Ϫ1 in the calcium-bound form.
Upon the addition of 150 M NS5806, the observed association rate for the fast and slow phase decreased ϳ2-fold; consequently, the association rate decreased but remained indepen-dent of calcium, k on fast ϭ 12 Ϯ 2 mM Ϫ1 s Ϫ1 and k on slow ϭ 1.1 Ϯ 0.8 mM Ϫ1 s Ϫ1 in the apo-state and k on fast ϭ 9 Ϯ 0.8 mM Ϫ1 s Ϫ1 and k on slow ϭ 1.1 Ϯ 0.3 mM Ϫ1 s Ϫ1 in the calcium-bound form. The pre-exponential factor associated with the slow phase increased from 0.08 in the apo-form to 0.33 in the calcium-bound form. The calculated dissociation rates decreased upon the addition of NS5806 and became partially calcium-independent with k off fast ϭ 24 Ϯ 3 ϫ 10 Ϫ3 s Ϫ1 and k off slow ϭ 0.03 Ϯ 0.01 ϫ 10 Ϫ3 s Ϫ1 in the apo-form and k off fast ϭ 17 Ϯ 2 ϫ 10 Ϫ3 s Ϫ1 and k off slow ϭ 0.02 Ϯ 0.01 ϫ 10 Ϫ3 s Ϫ1 in the presence of calcium.
The change in fluorescence intensity associated with binding of Kv4.3 site 1 to KChIP3-(161-256) were also best modeled with a double exponential function (Table 4). Similarly to KChIP3-(65-256), the association rate constants for the fast phase were mainly unaffected by binding of calcium. An increase in the pre-exponential factor associated with the slow phase (␣ 2 ) upon binding of calcium to KChIP3 was also observed. The calculated dissociation rates were affected by calcium binding to a larger extent than in the case of KChIP3-(65-256), with the fast phase decreasing from k off fast ϭ 5050 Ϯ 90 ϫ 10 Ϫ3 s Ϫ1 to k off fast ϭ 150 Ϯ 10 ϫ 10 Ϫ3 s Ϫ1 and the slow phase decreasing from k off slow ϭ 6.5 Ϯ 1.2 ϫ 10 Ϫ3 s Ϫ1 to k off slow ϭ 0.11 Ϯ 0.01 ϫ 10 Ϫ3 s Ϫ1 .
The addition of NS5806 resulted in a decrease in the association rate of the slow phase to k on slow ϭ 0.39 Ϯ 0.03 mM Ϫ1 s Ϫ1 in the apo-form and k on slow ϭ 0.42 Ϯ 0.04 mM Ϫ1 s Ϫ1 in the calciumbound form as well as a decrease in the dissociation rate of the fast and slow phases to k off fast ϭ 53 Ϯ 5 ϫ 10 Ϫ3 s Ϫ1 and k off slow ϭ 0.021 Ϯ 0.002 ϫ 10 Ϫ3 s Ϫ1 in the apo-form and k off fast ϭ 29 Ϯ 3 ϫ 10 Ϫ3 s Ϫ1 and k off slow ϭ 0.012 Ϯ 0.001 ϫ 10 Ϫ3 s Ϫ1 upon calcium binding. Unlike KChIP3-(65-256), the pre-exponential parameters for the slow phase were unaffected by the addition of NS5806 to apo-or Ca 2ϩ -bound KChIP3-(161-256). Together, these results indicate that the observed increase in affinity between Kv4.3 site 1 and KChIP3-(65-256) or KChIP3-(161-256) due to NS5806 binding can be attributed to a decrease in the dissociation rate of the site 1 peptide.
NS5806 Binds at the Hydrophobic Cavity Near EF-hand 4 -To identify the NS5806 binding site, docking simulations using AutoDock 4.2 algorithm were employed. Identical docking sites for NS5806 were identified on both constructs of KChIP3 (Fig.  5). The most prominent NS5806⅐KChIP3-(65-256) interactions are hydrophobic, between the nonpolar brominated phenyl ring on NS5806 and Phe-218, Tyr-174, and Ile-194 on KChIP3, whereas the more electrophilic fluorinated phenyl ring faces the solvent. We previously used a similar approach to KChIP3-(161-256) and found that the same hydrophobic cavity at the C terminus was the most favorable binding site (33). The fact that docking simulations identify the same docking site for 1,8-ANS and NS5806 correlate well with the 1,8-ANS displacement studies shown above. The direct role of Phe-218 and Tyr-174 in binding of NS5806 was confirmed by independent mutation of these residues to alanine. We observed that mutation of either residues results in a 2-3-fold decrease in the affinity of NS5806 in the calcium-bound state, whereas the affinity in the apostate for the Tyr-174 mutant increases about 2-fold to 17 Ϯ 0.8 M ( Table 3). The binding affinity in the presence of calcium of the Kv4.3 site 1 peptide decreased 3-fold for the Tyr-174 mutant while remaining unchanged for the Phe-218 mutant. In the presence of NS5806, KChIP3(Y174A) showed an increased affinity for the site 1 peptide in the apo-and calciumbound forms but still weaker than KChIP3-(65-256). On the other hand, KChIP3(F218A) in the presence of NS5806 showed lower affinity for the site 1 peptide in both apo and calcium states. Overall, these results support the docking simulation and the idea that these hydrophobic residues play an important role in NS5806 binding and its observed effect on Kv4.3 site binding. Furthermore, given that the binding site of NS5806 is located in the vicinity of the EF-hand 3 and 4, we tested whether drug association alters the KChIP3 interactions with Ca 2ϩ . Titration curves for Ca 2ϩ binding to KChIP3-(65-256)-FlAsH and NS5806⅐KChIP3-(65-256)-FlAsH are shown in Fig. 6; the resulting titration curves were best analyzed using a model that assumes two protein populations in the apo-state, each with at least two calcium binding sites (Equation 4). The recovered parameters show a dual population of KChIP3-(65-256)-FlAsH in the apo-state with the larger population (f ϭ 0.82) having a macroscopic binding constant K d ϭ 170 Ϯ 20 nM (Hill coefficient n a ϭ 1.5) and the minor population having a K d ϭ 10 Ϯ 2 M (Hill coefficient n b ϭ 1.1). The apparent dissociation constants are comparable to the calcium binding constants reported in previous studies, ranging from 0.1 to 10 M (14, 37), indicating that the addition of the amino acid sequence CCPGCC at the C terminus and labeling with the biarsenical fluorophore FlAsH did not alter the calcium affinity. Titration of calcium into KChIP3-(65-256)-FlAsH in the presence of 150 M NS5806 resulted in titrations curves that were also best fitted with a double Hill equation. The macroscopic binding constant recovered for the high affinity sites was K d ϭ 120 Ϯ 20 nM (Hill coefficient n a ϭ 1.1) with f ϭ 0.66 and a dissociation constant for low affinity sites of K d ϭ 21 Ϯ 4 M (Hill coefficient n b ϭ 1.4), indicating that the overall calcium affinity of KChIP3-(65-256) is marginally affected by binding of NS5806.

DISCUSSION
The precise mechanisms underlying the gating regulation of Kv4-mediated K ϩ currents are yet to be completely understood; however, it is widely accepted that these channels do not undergo open state inactivation involving the N terminus (38 -40). It has also been shown that the association of KChIPs at the N terminus of Kv4 channels impairs open-state inactivation (41). Electrophysiological studies support the idea that both the N and C termini of Kv4 channels interact to regulate the inactivation kinetics (40,42), opening the possibility that such interactions are modulated by binding of KChIPs at the N terminus (34,38). It is also proposed that the N terminus functions as a membrane transport control, which in the absence of KChIPs anchors Kv4 channels near the perinuclear region of the cell (43,44). The association of KChIPs at the T1 domain of Kv4 results in translocation to the membrane and a concomitant increase in total K ϩ current. In addition, the presence of KChIPs is necessary for recovery of K ϩ current with similar biophysical properties as those found in vivo.
Recent reports by Calloe (17) on the I TO -activating properties of a novel diphenyl urea compound (NS5806) whose activ-  -(161-256) (b). The solid line indicates the distance between tryptophan 169 and the dibromophenyl ring of NS5806. Initial NMR structures of KChIP3-(65-256) (PDB entry 2JUL) and KChIP3-(161-256) (PDB entry 2E6W) were used as docking macromolecules; EF-1 is shown in green; EF-2 is in orange; EF-3 is in blue; EF-4 is in red; NS5806 is shown as a ball-and-stick model. c, mutation of Tyr-174 or Phe-218 to alanine completely abolishes the affinity enhancement observed in the absence of calcium upon binding of NS5806. Conditions are as described in Fig. 4a.  NOVEMBER 14, 2014 • VOLUME 289 • NUMBER 46 ity depends on the presence of KChIPs has highlighted the role of these auxiliary proteins in regulation of I TO currents . This drug was shown to reverse the effect of induced heart failure in canine models (17,19). On the other hand, Witzel et al. (11) hypothesized that the observed current inhibition in hippocampal neuronal cells was likely due to a higher concentration of NS5806 used (20 M) compared with the concentration used in canine cardiomyocytes. Indeed, at NS5806 concentrations Ͼ100 M a reversal of activation was observed in HEK293 cells. This highlights the possibility of multiple binding sites of NS5806 on the Kv4-auxiliary subunit complex. Nonetheless, NS5806 still resulted in a slowing of inactivation and accelerated recovery from inactivation in both cardiac and neuronal cells at concentrations below 20 M.

Modulation of Kv4.3 and KChIP3 Interactions by NS5806
Given the dependence of K ϩ current modulation by NS5806 on the presence of KChIPs, in this report we investigated whether KChIPs could be a target binding partner for NS5806. Titration data show that both the apo-and calcium-bound forms of KChIP3-(65-256) were quenched by NS5806 with an affinity 5-fold lower in the apo-form. The hydrophobic nature of NS5806 supports the idea that the binding site on KChIP3 is located in a solvent-restricted hydrophobic cavity and that calcium binding to KChIP3 increases accessibility of the drug to this cavity. Moreover, the similar dissociation constants recovered for both KChIP3 constructs indicate that in the calciumbound and apo-state of the protein the binding site of NS5806 is located at the C terminus. Interestingly, the dissociation constants recovered for NS5806 binding to Ca 2ϩ KChIP3-(65-256) are very similar to the EC 50 (Fig. 2a, inset). These results are in agreement with fluorescence lifetime data which show that the addition of NS5806 results in a concomitant decrease in the pre-exponential parameter ␣ 2 associated with 1,8-ANS bound at the solventrestricted cavity (Fig. 2c), whereas the pre-exponential parameter that corresponds to 1,8-ANS bound to a solvent-exposed site with emission maximum at 485 nm remains constant. These results suggest that 1,8-ANS binds at a physiologically active hydrophobic cavity on KChIP3 and that displacement of 1,8-ANS could be used as a fluorescence-based high throughput method for the discovery of new drugs that bind at this hydrophobic cavity and could induce similar effects to that observed for NS5806. Indeed, an analogous 1,8-ANS screening assay has been previously reported for other drug targets (45).
Isothermal calorimetric titrations of NS5806 into KChIP3 further confirm the strong association between NS5806 and KChIP3. The recovered thermodynamic parameters show that both enthalpic and entropic contributions stabilize the complex formation, with the enthalpic contribution becoming dominant in the KChIP3-(161-256) construct. The large enthalpy contribution is in contrast to the idea that the binding site of NS5806 on KChIP3 is mainly hydrophobic (entropydriven binding); however, it is likely that binding of NS5806 triggers larger exothermic structural rearrangements on the protein.
Characterization of the interaction of KChIP3 and peptides homologous to Kv4.3-(2-22) and Kv4.3-(70 -90) indicate that the interactions between site 1/site 2 and KChIP3 are modulated by calcium (Table 3). In the presence of calcium KChIP3-(65-256) showed a 26-fold increase in affinity for site 1, whereas KChIP3-(161-256) showed an 18-fold increase compared with the apo-form. Furthermore, the dissociation constant recovered for site 1 binding to KChIP3-(161-256) in the presence of calcium was about 10-fold weaker than that of site 1 binding to KChIP3-(65-256). The determined dissociation constants can be utilized to calculate the role of the N-and C-terminal domains of KChIP3 on binding of site 1 of Kv4.3. The energetic contribution of the N and C termini of KChIP3 to the association with site 1 calculated from the dissociation constants in the calcium-bound state are Ϫ1.3 and Ϫ6.1 kcal mol Ϫ1 , respectively. The larger contribution of the C terminus highlights the role of this domain on KChIP3 as mediator in protein-protein interaction. Interestingly, the Y134E mutation on KChIP1 (Y174E on KChIP3) has also been shown to completely abolish current modulation by KChIP1 (46). However, the contribution of the N terminus hydrophobic cavity of KChIPs is necessary for Kv4 translocation to the membrane and KChIP-Kv4 complex formation (47,48). This is in agreement with structural studies, where the N terminus of Kv4.3 was determined to bind across the hydrophobic face of KChIP1, contacting residues spanning both the N and C termini of KChIP1 (32,36). The dissociation constant for site 2 binding to KChIP3-(65-256) was also dependent on calcium, with an ϳ50-fold increase in affinity but still weaker than the site 1 interaction. The fact that calcium binding at the C terminus EF-3 and EF-4 of KChIP3 modulates binding of site 2 is remarkable, suggesting that the structural changes associated with the calcium binding to the C-terminal domain are propagated into the N-terminal domain of the protein. However, it is also possible that the calcium-induced dimerization of KChIP3 is responsible for the increase in affinity.
The addition of saturating amounts of NS5806 completely abolished the calcium dependence of site 1 association with KChIP3 as well as further increased its affinity. These results are surprising given that the experiments above support the idea that NS5806 and site 1 share the same binding site at the C terminus of KChIP3. Furthermore, a clear dose-dependent increase in KChIP3-site 1 association was observed upon the addition of NS5806 in the absence of calcium (K d ϭ 21 Ϯ 1 M), but no further increase was observed upon the addition of excess calcium. On the other hand, at saturating amounts of calcium, the addition of NS5806 induced a further increase in anisotropy, suggesting that in the presence of NS5806 the KChIP3-site 1 complex adopts an altered tertiary structure, as evident from the larger anisotropy upon binding in the presence of NS5806 (Fig. 4b). These results are comparable with those obtained for the Kv4.3(1-143)-KChIP1 complex using size exclusion chromatography in the presence of the diphenylurea compound CL-888 (49). A similar effect due to the presence of NS5806 for site 1 bound to KChIP3-(161-256) on anisotropy was observed, indicating that structural rearrangement is localized on the C terminus domain of KChIP3.
A better understanding of the mechanism by which NS5806 modulates the affinity of site 1 peptide binding to KChIP3 construct is gained from the kinetics associated with binding ( Table  4). The biphasic nature of the binding kinetics is potentially due to the presence of KChIP3 proteins that populate partially distinct conformations or oligomerization state. Overall, these results revealed that saturating amounts of NS5806 decreased the rate of dissociation between KChIP3 and site 1 of Kv4.3 independent on the presence of calcium. A possible explanation for the stabilization of the protein peptide complex is that binding of NS5806 at the C-terminal cavity destabilizes this domain and increases the accessibility of the hydrophobic residues to interact with hydrophobic residues on site 1. Also, the fact that the affinity of KChIP3 in the apo-form for NS5806 is lower than in the presence of calcium indicates that there is a larger energy barrier for access to the hydrophobic cavity at the C terminus. However, once NS5806 binds at the hydrophobic cavity in the apo-form, the resulting structural changes are comparable to those present in the calcium-bound form. This results in similar affinity and dissociation constants for the N terminus of Kv4.3. On the other hand, the observed decrease in dissociation rate in the presence of NS5806 seems to indicate that the KChIP3-NS5806 populates a conformation that is different from that of the apo-form and calcium-bound KChIP3 and forms a more stable complex with site 1 of Kv4.3.
Computational simulation allowed us to pinpoint the potential docking site for NS5806 as being at a hydrophobic cavity between EF-3, EF-4, and the H10 helix. This docking site is also in good agreement with the distances from Trp-169 determined using resonance energy transfer and with the displacement studies. However, previous docking studies of a similar diphenyl-urea compound (CL-888) identified a site on the hydrophilic surface of KChIP1 (49). The proposal that this is the docking site for NS5806 is not supported by the fact that similar dissociation constants were recovered for KChIP3-(65-256) and KChIP3-(161-256) even though the C-terminal construct lacks half of the residues involved in the proposed docking site for CL-888. We also conducted a docking simulation with NS5806 where we limited the docking grid to cover only the proposed binding site at the hydrophilic surface. The predicted association energy of this site was 4 kcal mol Ϫ1 weaker than the predicted energy for the docking at the hydrophobic cavity near EF-3 and EF-4. Indeed, the role of Tyr-174 and Phe-218 identified in docking simulation as being involved in stabilization of NS5806 and Kv4.3 site 1 peptide interactions is supported by site-directed mutagenesis studies, which show that mutation of either residue to alanine completely abolishes the effect of NS5806 (Fig. 5c). We also observed that a KChIP3(F252A) mutant readily forms aggregates (data not shown), supporting the idea that this residue plays an important role in the stability of the protein.
Therefore, we propose that binding of NS5806 at the C-terminal hydrophobic cavity results in an increase in flexibility of the H10 helix that is likely facilitated by displacement of the hydrophobic residues on this helix. This is supported by molecular dynamic simulations which show that in the absence of NS5806 the aromatic residue Phe-252 on the H10 helix is stacked between Tyr-174 and Phe-218, inherently reducing the flexibility of this helix (Fig. 7). In contrast, simulations in the presence of NS5806 show a displacement of Phe-252 residue and concomitant motion of the H10 helix to accommodate NS5806 in the hydrophobic cavity. Furthermore, simulation of the Kv4.3⅐KChIP3 complex show that Trp-19 of Kv4.3 is positioned in an identical stacking position between Tyr-174 and Phe-218 as that found for Phe-252 and NS5806, whereas in the presence of NS5806 the cavity at the C terminus of KChIP3 expands to accommodate both NS5806 and the hydrophobic N terminus of Kv4.3 while pushing Trp-19 deeper into the hydrophobic cavity (Fig. 7). The motion of Trp-19 into the hydrophobic cavity of the KChIP3 C terminus may contribute to the slower rate of dissociation of the hydrophobic site 1 peptide in the presence of NS5806. The structural rearrangement observed using molecular dynamics is consistent with the idea that binding of NS5806 induces structural rearrangements in the Kv4.3⅐KChIP3 protein complex. Whether these structural changes are isolated on KChIP3 and the N terminus of Kv4.3 or they propagate across the T1 domain remains to be determined. Previous studies favor the idea that such structural changes could propagate toward the K ϩ channel inner vestibule near the pore and modulate the K ϩ currents (50,51).
The results presented here could potentially explain the observed increase in K ϩ current and the slowing of inactivation in the presence of NS5806 in cardiac, neuronal, and heterologous expression studies. The elimination of the calcium dependence of association between KChIP3 and site 1 of Kv4.3 as well as the increase in affinity in the presence of NS5806 support the idea that the observed increase in K ϩ current amplitude is due to an enhanced translocation of Kv4.3⅐KChIP3 protein complexes to the cell membrane induced by the presence of NS5806. We further hypothesize that the resulting decrease in inactivation kinetics is due to the decrease in accessibility of the N terminus of Kv4.3 channel to interact with the cytoplasmic C terminus of Kv4. Indeed, this interaction has been shown to be necessary for modulation of gating currents of Kv4.2 (35). The fact that the dissociation constant was also reduced in the presence of calcium also highlights the potential role of this ion in K ϩ current regulation. In support of the role of calcium in regulating K ϩ current are studies which have shown that in rat stellate cells a protein complex is formed involving the voltage-gated calcium channel Cav3, the potassium channel Kv4.2, and the auxiliary protein KChIP3 (52).