|Home | About | Journals | Submit | Contact Us | Français|
Brain IA and cardiac Ito currents arise from complexes containing Kv4 voltage-gated potassium channels and cytoplasmic calcium-sensor proteins (KChIPs). Here, we present X-ray crystallographic and small-angle X-ray scattering data that show that the KChIP1–Kv4.3 N-terminal cytoplasmic domain complex is a cross-shaped octamer bearing two principal interaction sites. Site 1 comprises interactions between a unique Kv4 channel N-terminal hydrophobic segment and a hydrophobic pocket formed by displacement of the KChIP H10 helix. Site 2 comprises interactions between a T1 assembly domain loop and the KChIP H2 helix. Functional and biochemical studies indicate that site 1 influences channel trafficking, whereas site 2 affects channel gating, and that calcium binding is intimately linked to KChIP folding and complex formation. Together, the data resolve how Kv4 channels and KChIPs interact and provide a framework for understanding how KChIPs modulate Kv4 function.
In biological systems, electrical information is encoded and processed by changes in action-potential timing, duration, frequency, waveform and number1. Modulation of voltage-gated potassium channels is central to these events and affects heart rate, sensory transduction and cognition. The ion transport2 and voltage-dependent gating properties3 of the potassium channel α subunits that form the ion conduction pathway are well characterized. Although α subunits form the pore, many channels function as complexes that require cytoplasmic and transmembrane auxiliary subunits4. To date, little is known regarding the way in which such components bind α subunits and modulate channel action. Understanding the interplay between regulatory components and pore-forming domains is crucial for unraveling how modulatory signals4 and homeostatic mechanisms5 allow excitable cells to sense and respond to environmental cues.
Two potassium currents, IA (ref. 6) and Ito (ref. 7), exert strong control over neuronal and cardiac excitability, respectively, and provide clear examples of the importance of auxiliary subunits for tuning properties of α subunits. Both currents arise from complexes of Kv4 voltage-gated potassium channel subunits with a family of soluble calcium-sensor ‘Kv-channel interacting proteins’, KChIPs6,8–10. A related protein, frequenin, has effects on Kv4s that are similar to those of KChIPs10,11. KChIPs and frequenin belong to the large family of neuronal calcium-sensor proteins8,12. Although the presence of calcium-sensor proteins as essential components of the IA and Ito complexes suggests that calcium may be important for channel regulation, no clear physiological demonstration of this link has been given.
KChIPs have profound effects on the properties of Kv4 α subunits. Their coexpression with Kv4s increases current density, changes channel inactivation kinetics and speeds recovery from inactivation8,13–16. The KChIP-Kv4 interaction requires the Kv4 N-terminal cytoplasmic domain, which comprises a ~40-residue N-terminal hydrophobic segment that is unique to Kv4 channels8,13,15, and the Kv channel assembly domain, T1 (ref. 17). KChIP–Kv4 complexes have deep evolutionary roots. Genes for each member are found in chordates as primitive as Ciona intestinalis18. Cell-biological studies suggest that interactions between KChIPs and the N-terminal hydrophobic segment regulate channel trafficking to the plasma membrane13,15,19. Loss of KChIP2 function leads to enhanced arrhythmia susceptibility, underscoring the crucial nature of the KChIP-Kv4 interaction20.
The way KChIPs and Kv4 channels associate remains unresolved. Three studies have suggested mutually exclusive models. Biochemical work has shown that, in addition to interactions with the N-terminal hydrophobic segment that precedes the T1 domain, KChIP interactions with a loop on the T1 cytoplasmic face, the ‘docking loop’, are important17. These data led to the suggestion that KChIPs dock underneath the T1 complex in a way that resembles the Kv1 T1 domain–Kvβ subunit interaction, where two tetrameric structures dock in series21. In contrast, an interpretation of low-resolution (21 Å) negative-stain images of complete KChIP2–Kv4.2 complexes has suggested a model where the KChIPs are not underneath the T1 domains but rather form a square-shaped planar structure in which the KChIPs run parallel to the T1 tetramer sides22. The third model originates from studies of an engineered KChIP1 in which the first 30 residues of the Kv4.2 hydrophobic N terminus were fused directly to the C-terminal end of the final KChIP1 helix, H10. The chimera formed dimers in solution. From these data, the authors concluded that KChIP dimers associate with the T1 domain in a way that breaks with the four-fold symmetry of the α subunit23.
Here, we report crystallographic and solution experiments that demonstrate that the KChIP1–Kv4.3 T1 complex is a cross-shaped octamer with four-fold rotational symmetry that matches the pore-forming domains21. The structure resolves the question of how these proteins interact and defines two sites of KChIP-Kv4 interactions that have distinct functional roles in channel trafficking and gating.
We used a coexpression strategy to define which lengths and combinations of Kv4 N-terminal cytoplasmic domains and KChIP isoforms formed soluble, well-behaved complexes in Escherichia coli. We crystallized a complex of the rat Kv4.3 (also called Kcnd3) N-terminal cytoplasmic domain (residues 1–143) and human KChIP1 (also called KCNIP1; residues 37–216) containing the conserved, functional KChIP core24. Initial crystals diffracted poorly and could not be improved. Thus, we used surface entropy reduction and converted selected surface lysine residues to alanine25. A double mutant (K160A K167A) in the nonconserved KChIP1 EF3-EF4 linker yielded crystals that diffracted X-rays to 3.35 Å. Electrophysiological examination showed that KChIP1(K160A K167A)–Kv4.3 channel complexes were indistinguishable from wild-type complexes (data not shown).
The structure of the KChIP1–Kv4.3 T1 complex shows that the assembly forms a cross-shaped octamer having the T1 tetramer at the center (Fig. 1a, chains A–D) and individual KChIPs extending radially (Fig. 1a, chains E–H). The asymmetric unit has two octameric complexes that are apposed on the cytoplasmic T1 faces (Fig. 1b). T1 shows few differences from the isolated Kv4.3 T1 structure17 (r.m.s. deviation = 0.64 Å2 over 408 Cα positions). In contrast, KChIP1 shows substantial differences (r.m.s. deviation = 4.03 Å2 over 179 Cα positions) from isolated KChIP1 (ref. 17).
The structure reveals two main interaction sites. Site 1 buries ~2,100 Å2 total surface area between residues 3 and 21 of a conserved hydrophobic segment that is on the N-terminal side of T1 (called T1N; Fig. 2a) and a large hydrophobic pocket (28 Å long, 12 Å deep, 10 Å wide) formed by KChIP1 (Fig. 2b,c). T1N contains residues essential for KChIP-Kv4 interaction8,13 and has two parts: an α-helix, residues 4–17 (the T1N helix), and a loop, residues 18–39, lacking regular secondary structure (the T1N linker). Site 2 comprises contacts between the KChIP1 H2 helix and the T1 domain docking loop that bury ~900 Å2 (Fig. 2d). Each KChIP1 uses site 1 interactions to bind T1N from one Kv4.3 monomer but makes site 2 interactions with the adjacent T1 monomer (Fig. 1a). The contacts with two T1 subunits seen in this arrangement are consistent with the report that, in full-length channels, KChIP binding can overcome T1 assembly defects26 and may endow the KChIP–Kv4 T1 complex with cooperative properties that affect gating.
As in other KChIP structures17,23, only EF3 and EF4 of the EF-hand calcium-binding loop sequences are in the calcium-bound state (Fig. 2e and Supplementary Fig. 1 online). EF1 lacks key calcium-binding residues and is empty. EF2 is unoccupied and is distorted by the binding of the T1N helix (Supplementary Fig. 1).
Comparison with the isolated KChIP1 structure17 indicates that a large conformational change creates site 1. The H10 helix moves to the rim of the wide, hydrophobic groove (Fig. 2e). This conformation is comparable to the frequenin structure (Supplementary Fig. 1), where the hydrophobic pocket is occupied by polyethylene glycol molecules from the crystallization solution27. The similar conformational change suggests that H10 displacement from its position as a cover over the site 1 hydrophobic pocket is a general mechanism by which KChIPs and frequenin bind target proteins.
The T1N helix interacts with the KChIP1 hydrophobic pocket in a parallel orientation. This orientation is opposite to that observed in a fusion protein in which the first 30 N-terminal residues of Kv4.2 are fused covalently to the KChIP1 H10 helix C terminus23, where the identical Kv4.2 T1N residues do not induce a conformational change in the site 1 pocket (Supplementary Fig. 1). Three observations support our T1N register and orientation assignment. First, we observed strong electron density for T1N Trp8, Phe11 and Trp19 (Supplementary Fig. 2 online). Second, T1N residues 21–39 have sufficiently clear density in noncrystallographic symmetry–averaged maps to define the T1N connectivity with T1 (Supplementary Fig. 2). Third, the Sed atom of Met20 (Supplementary Fig. 2) has clear density (4.5 σ) at the expected position in noncrystallographic symmetry–averaged anomalous difference maps generated using the amplitudes from a crystal of the selenomethionine (SeMet)-substituted complex. These results unambiguously support our assigned helix direction and register and suggest that the previously reported structure is a consequence of the unnatural fusion.
T1N residues 7–21, which are strongly conserved (Fig. 2a), make most of the interactions with the KChIP1 hydrophobic pocket. The three aromatic residues, Trp8, Phe11 and Trp19, each of which is important for KChIP-Kv4 interaction and current modulation23,28, fit into hydrophobic crevices within the site 1 pocket (Fig. 2c). Trp8 contacts KChIP1 Phe60, Val69, Phe74, Ile77 and Phe111. Phe11 contacts an adjacent crevice composed of KChIP1 Phe98, Phe111, Ala114, Leu115, Leu118 and Trp129. Trp19 interacts with KChIP1 Tyr134, Ile150, Ile154 and Tyr155 as well as T1N Met20. The short stretch of conserved T1N alanines (Ala12, Ala14, Ala15 and Ala16; Fig. 2a) allow the T1N helix to fit tightly into the KChIP1 pocket.
The contacts between T1N and KChIP1 are concordant with previous studies. Deletion analyses have shown that truncation of Kv4.2 residues 2–20 (ref. 13), and the shorter truncation of residues 7–11, but not 2–6 (ref. 17), greatly reduce interactions of KChIP with the Kv4.2 cytoplasmic N-terminal domain. Furthermore, identification by lysine scanning mutagenesis of Ile17, Gly18 and Met20 as residues crucial for interaction28 also fits well with the intimate interaction of these positions with the hydrophobic pocket.
Site 2 buries 900 Å2 in an interface of mixed hydrophilic and hydrophobic character in which the KChIP1 H2 helix contacts the α2 T1 helix and the docking loop on the cytoplasmic face of the T1 complex (Fig. 2d). Kv4.3 Phe73 contributes the major hydrophobic component and is involved in aromatic stacking interactions with KChIP1 Tyr57. The Kv4.3 Asp78 side chain is positioned to form a salt bridge with KChIP1 Arg51. The Kv4.3 Glu72 and Phe75 main chain oxygens appear within hydrogen-bonding distance of KChIP1 Gln54. Finally, the Kv4.3 Glu70 side chain is well situated to contact the side chains of KChIP1 Tyr57 and Lys61. Kv4.3 T1 docking loop residues Glu70, Phe73 and Asp78, which are identical to residues in Kv4.2 known to be crucial for KChIP-Kv4.2 interactions17,28, reside squarely in the site 2 interface and support the relevance of the structure with respect to the full-length channel.
We used small-angle X-ray scattering (SAXS)29 to probe the KChIP137–216–Kv4.31–143 complex structure in solution and test whether the flat, cruciform octamer is the preferred arrangement or a consequence of crystallization. SAXS provides an independent assessment of the overall shape of protein molecular envelopes at ~10- to 20-Å resolution29.
Two different analyses, the Guinier approximation and GNOM30, yielded similar values for the radius of gyration (Rg = 42.8 ± 0.7 Å and 43.5 ± 0.4 Å, respectively). The scattering element distribution described by the P(r) function defines the longest axis in the complex (Dmax = 155 ± 5 Å; Supplementary Fig. 3 online). These Rg and Dmax values agree with those calculated by CRYSOL31 for our crystal structure of the octamer (Rg = 42.6 Å and Dmax = 150 ± 5 Å).
We used GASBOR32 to calculate low-resolution ab initio models of three-dimensional arrangements of scattering centers that reproduce the SAXS data and provide the shape of the molecular envelope of the complex (Supplementary Fig. 3). Because this inverse scattering problem has no unique solution, we performed ten independent runs for three different symmetries (P1, P2 and P4) and compared and averaged the models from each using DAMAVER33. Calculations using P1 or P2 symmetry showed wide variations in the distributions of dummy residues that did not converge. In contrast, the models obtained using P4 symmetry converged and fit experimental scattering data well (χ2 = 3.46). The resulting model is cross-shaped (Supplementary Fig. 3) and strongly resembles our crystallographically determined complex. Comparison of the calculated scattering profile from our crystallographic structure using CRYSOL31 and the measured data shows excellent agreement (χ2 = 3.03) that is also evident from superposition of the octamer structure and SAXS model (Fig. 3a).
A flat, four-fold symmetric arrangement of the KChIP–Kv4 T1 complex has been proposed from low-resolution (21 Å) negative-stain images of KChIP2–Kv4.2 (ref. 22). Superposition of a square-shaped complex similar to that proposed in ref. 22, in which the KChIPs run laterally to the T1 domains, shows poor overlap with the SAXS model (Fig. 3b). This square-shaped arrangement has an inferior correlation to the SAXS data, compared with the crystallographic cross-shaped octamer, when the scattering curves are compared using CRYSOL (χ2 = 7.06) and has substantially smaller Rg and Dmax values (Rg = 36.1 Å, Dmax = 110 ± 5 Å) than the experimental values (see above). These great differences exclude the possibility that there is an appreciable population of the square-shaped arrangement in solution, eliminate the possibility that the cross-shaped character of the ab initio SAXS model results solely from the use of P4 symmetry and provide a noteworthy example of the power of SAXS to vet competing models of macromolecular complex architecture. Furthermore, dynamic light-scattering experiments, as well as comparison of the expected SAXS profiles of the individual components and the unit cell hexadecameric arrangement, firmly support the conclusion that the complex is an octamer (Supplementary Fig. 3). Together, these data show that the shape of complex has not been forced by crystallization conditions and strongly suggest that the cruciform complex reflects the native KChIP1–Kv4.3 arrangement.
Because the structure defines a previously uncharacterized KChIP-T1 interface, site 2, we mutated key KChIP1 residues in this interface (Arg51, Gln54, Tyr57 and Lys61) to alanine to examine its functional role. Xenopus laevis oocyte two-electrode voltage clamp experiments revealed that Y57A and K61A mutations did not alter surface expression but clearly perturbed the inactivation kinetics of the channel complex, such that the mutants resembled channels lacking KChIPs (Fig. 4a,b and Table 1). R51A and Q54A perturbed neither gating nor surface expression. The double mutant, Y57A K61A, had inactivation and recovery-from-inactivation properties indistinguishable from Kv4.3 channels lacking KChIPs (Fig. 4b,c and Table 1) yet retained the potent effect on current enhancement (Fig. 4a and Table 1). Consistent with the smaller buried surface area at site 2 compared with site 1, none of the mutations caused substantial perturbation of binding to the HMT-Kv4.31–143 fusion protein (Fig. 4d). Together, these data suggest that site 1 and site 2 have two separate roles. Site 1 interactions seem to be largely responsible for KChIP-Kv4 association and for mediating channel trafficking effects, whereas site 2 interactions influence channel inactivation and recovery from inactivation.
In neuronal calcium-sensor proteins such as KChIPs, calcium-dependent conformational changes of the EF hands are thought to be important for function12. Simultaneous disablement of KChIP EF2, EF3 and EF4 eliminates modulation of Kv4 currents8. Because the importance of calcium binding to each functional KChIP EF hand was unknown, we mutated the first calcium-coordinating aspartate of each one singly (EF2, D99A; EF3, D135A; EF4, D183A), in pairs (EF2 EF3, D99A D135A; EF2 EF4, D99A D183A; EF3 EF4, D135A D183A) and as a triple mutant (EF2 EF3 EF4, D99A D135A D183A) and tested the mutants’ ability to bind the HMT-Kv4.31–143 fusion protein. The single mutants and the EF2 EF3 and EF2 EF4 double mutants did not measurably impair binding. In contrast, the EF3 EF4 and EF2 EF3 EF4 mutants were clearly defective (Fig. 5a).
To understand why certain KChIP1 mutants did not bind, we expressed and purified KChIP1 and each mutant individually to characterize their properties. KChIP1 and the EF2, EF4 and EF2 EF4 mutants each eluted from a gel-filtration column in a single peak corresponding to a monomer (Supplementary Fig. 4 online). In contrast, the EF3 and EF2 EF3 mutants eluted in several peaks corresponding to a mixture of monomer and aggregated protein. Both EF3 EF4 and EF2 EF3 EF4, the mutants that did not bind the intracellular N-terminal Kv4.3 domain, eluted as a peak that had a substantially higher molecular weight than wild-type KChIP1 (Supplementary Fig. 4). SDS-PAGE analysis showed apparently stoichiometric amounts of a protein that we identified by western blotting as the E. coli chaperone DnaK (Supplementary Fig. 4). The tight association of these chaperone–KChIP EF-hand mutant complexes suggests that KChIP folding and calcium binding are intimately linked.
We used spectroscopic experiments to characterize further the structural consequences of KChIP-calcium interactions. Circular dichroism experiments show that reduction of free calcium concentration causes KChIP1 to become less helical (Fig. 5b). Dynamic light-scattering experiments indicate that this KChIP1 conformational change is concomitant with the conversion from a well-behaved monodisperse protein into a form that aggregates (Supplementary Fig. 4). We obtained similar results for the EF2, EF4 and EF2 EF4 mutants (data not shown). Together with the biochemical data, these results demonstrate that calcium binding is required for proper KChIP folding and is essential for binding to Kv4 channels.
The observation that EF2 EF3 EF4 binds poorly to the Kv4 N-terminal cytoplasmic domain disagrees with reports using full-length channels8, where interactions with the Kv4 C-terminal cytoplasmic domain28 and localization of full-length KChIP1 to the membrane by an N-terminal myristoylation34 may also contribute. To discover whether these factors may compensate the defects caused by the EF-hand mutations and to gain further insight into the role of calcium, we pursued functional characterization of full-length Kv4.3 with KChIP1 EF-hand mutants in Xenopus oocytes. As reported8, coexpression of Kv4.3 and KChIP1 increases current amplitude, slows inactivation and speeds recovery from inactivation (Table 1). Current amplitudes and kinetic properties of Kv4.3 coexpressed with the EF2, EF4 or EF2 EF4 mutant were similar to those upon coexpression with wild-type KChIP1. The two mutants with poor biochemical properties (EF3 and EF2 EF3) had concentration-dependent effects (Fig. 5c–e). At a 1.5:1 Kv4.3/KChIP1 mRNA molar ratio, EF3 and EF2 EF3 mutants behave similarly to wild-type KChIP1. Reduction in the relative amount of KChIP1 (3.5:1 Kv4.3/KChIP1) spared the effects of wild-type KChIP1 on kinetics but made the EF3 and EF2 EF3 mutants ineffective at modulation. In agreement with their severely impaired binding ability, EF3 EF4 and EF2 EF3 EF4 mutants had no effect on current peak amplitudes or kinetics and produced channels similar to Kv4.3 alone (Fig. 5c–e). Together, these results indicate that KChIP calcium binding is important for Kv4.3 binding and channel modulation.
Our work provides the first high-resolution description of KChIP-Kv4 association. The identification of site 1 and site 2 as principal points of contact agrees with previous mutagenesis data that indicated the importance of both the T1N segment13,17 and the T1 docking loop17,28. The planar, cross-shaped octameric structure of the complex clearly differs from the previously suggested structural models17,22,23 and resolves the question of how KChIPs and Kv4 N-terminal cytoplasmic domains interact. Superposition of the KChIP1–Kv4.3 complex and the homologous Kv1.2 T1 domains in the Kv1.2 channel structure21 (Fig. 6a) suggests a notable offset between the positions of the KChIPs and the voltage sensors (Fig. 6b) and a resemblance to the overall shape of the particles seen in KChIP2–Kv4.2 complex22.
Each Kv4 subunit has a ~150- to 200-residue C-terminal cytoplasmic domain. Topological considerations require that this domain exits the transmembrane domain near the four-fold central axis of the pore and near T1 (ref. 35), where it may make specific interactions with T1 and the KChIPs. Negative-stain images of KChIP2–Kv4.2 show pillars of density at the complex periphery that could represent part of the α subunit's C-terminal domain22, and functional studies indicate that C-terminal truncations of the a subunit may interfere with KChIP-Kv4 assembly28. Thus, other points of direct KChIP-Kv4 physical interaction may contribute to assembly and gating.
Fast inactivation of Kv4 channels seems to be fundamentally different from the Shaker-type ‘ball-and-chain’ mechanism, in which the N-terminal cytoplasmic segment of the protein is thought to enter and block the channel pore1. Kv4 inactivation is thought to arise from the concerted action of the N- and C-terminal cytoplasmic domains36. Although the Kv4 N terminus can influence inactivation in the absence of KChIPs37,38, the intimate association of T1N and KChIP excludes the possibility that KChIP–Kv4 inactivation occurs by a ball-and-chain mechanism and supports the idea that T1 conformational changes are coupled to Kv channel gating35,39,40. Exactly how KChIPs, T1 and the α subunit's transmembrane segments are mechanistically coupled remains to be explained. In light of the primacy of site 2 for modulating inactivation and recovery from inactivation, the interaction of KChIP with more than one T1 subunit offers a likely means by which KChIPs could influence gating.
Different KChIP isoforms affect α-subunit kinetics in distinct ways41–43. Most KChIP variability resides in the nonconserved N-terminal region absent from our structure. It seems likely that this KChIP portion is responsible for the varied functional effects41–43. A truncated KChIP (KChIP2d) lacking the entire protein except for EF4 has been reported to affect channel properties16,44. Because KChIP2d lacks the site 1 hydrophobic cleft and all of site 2, the reported functional effects are unlikely to be related to the function of intact KChIPs.
Kvβ cytoplasmic subunits also bind T1 and affect channel function21,45–47. These subunits can bind Kv4s48,49, and simultaneous overexpression of KChIP2 and Kvβ3 produces channels with unique properties50; however, it is unclear whether native Kv4s associate simultaneously with KChIPs and Kvβs. Comparison of KChIP1–Kv4.3 T1 and Kv1.2–Kvβ21 complexes indicates that the two accessory subunits bind opposite faces of the T1 docking loop and suggests that formation of a Kv4–KChIP–Kvβ complex is possible (Fig. 6c,d). Superposition of T1 in the KChIP–Kv4.3 T1 complex with T1 in the Kvβ–Kv1.2 complex21 (Fig. 6d) shows remarkably little steric clash between the structures, except for two side chain pairs whose clashes could easily be relieved by rotamer changes. Thus, the T1 docking loop may serve as a site for KChIP–Kv4 channels to engage Kvβ or other modulatory proteins that could further tune channel properties. The potential serial arrangement (Fig. 6c) suggests an organizational principle for ion-channel complex formation: the membrane layer contains the pore and voltage sensors, whereas the membrane-proximal layer contains assembly information and components that regulate gating and trafficking; this membrane-proximal layer may in turn accept input from a distal layer comprising other regulatory cytoplasmic factors.
An important question is whether there is an in vivo link between calcium levels and Kv4 function. Cell-biological studies have demonstrated that without KChIPs, Kv4 channels do not reach the plasma membrane efficiently and are largely retained intracellularly15,51. Coexpression with KChIP1, KChIP2 or KChIP3 causes a marked redistribution of channels to the plasma membrane that requires functional EF hands51. Our data show that in the absence of calcium, KChIP1 has a strong tendency to unfold and aggregate. Binding to the Kv4 N-terminal cytoplasmic domain and modulation of channels are also impaired. Thus, without calcium, KChIPs are unable to bind and modulate Kv4 channels. These results suggest that intracellular calcium changes could alter the quantity of fully assembled KChIP–Kv4 complexes in the membrane. Whether such regulation is acute (milliseconds to seconds) or on a longer timescale (minutes to hours) remains to be determined. Our demonstration of the importance of calcium for KChIP folding and Kv4 association suggests that the latter may be crucial.
In cardiac ventricles, the Ito transmural gradient seems to be determined by KChIP2 expression levels52,53. Decreases in Ito are associated with cardiac arrhythmias7,20 and the cardiac action-potential prolongation that is associated with cardiac hypertrophy and heart failure7,54. IA has a nonuniform density distribution in dendrites and is important in action-potential backpropagation6. Changes in cardiac Ito and neuronal IA may reflect the ability of Kv4 channel complexes to sense and respond to environmental cues that are a part of normal or pathophysiological processes. In this respect, the link between calcium-mediated folding of KChIPs, binding to Kv4 α subunits and increased surface expression may be important. Furthermore, small molecules that could alter Ito or IA channel activity by modulating KChIP-Kv4 interactions could tune the excitability of a variety of tissues55. The KChIP1–Kv4.3 T1 structure provides the necessary molecular architecture for detailed investigation of how persistent changes in calcium levels might affect channel regulation and may prove an important guide in the development of new channel pharmacopeias.
Rat Kv4.31–143 (GenBank L48619) was cloned into a modified pET28 vector containing a His6 tag, maltose-binding protein (MBP) and a tobacco etch virus cleavage site on the N terminus of the protein sequence (pET28-HMT)56. Human KChIP1 (residues 37–216; IMAGE clone) was cloned into a pEGST vector lacking the glutathione S-transferase (GST) tag57. Both vectors were cotransformed into E. coli BL21(DE3)pLysS. Protein expression and purification details can be found in Supplementary Methods online.
Crystals of native KChIP137–216(K160A K167A)–Kv4.31–143 complexes and SeMet-substituted complexes were obtained at room temperature by hanging drop vapor diffusion and flash-frozen in mother liquor supplemented with 20% (v/v) glycerol. Data were collected at Lawrence Berkeley National Laboratories Advanced Light Source (ALS) beamline 8.3.1 (Table 2). Details of the crystallization, structure solution and refinement procedures can be found in Supplementary Methods.
Circular dichroism spectra of KChIP137–216 and EF mutants at 20 μM in 250 mM KCl and 10 mM HEPES (pH 7.4), in the presence of 1 mM Ca2+ and 0–10 mM EGTA, were recorded on an Aviv 215 circular dichroism spectrometer using a 1-mm-path-length quartz cell at 20 °C. Dynamic light-scattering experiments used a DynaPro instrument (Wyatt Technologies).
Constructs for electrophysiology were subcloned into a pGEMHE derivative58. RNA transcripts (Ampliscribe, Epicentre Technology) were injected into Xenopus oocytes as described58. Kv4.3 and KChIP1 mRNA final concentrations were 12.5 ng μl–1 and 5 ng μl–1, respectively, unless otherwise indicated. Two-electrode voltage-clamp experiments were performed 1 or 2 d after injection using CLAMPLEX 184.108.40.206, previously described protocols59 and a GeneClamp 500B (Axon Instruments) amplifier controlled by a computer with a 1,200-MHz processor (Celeron, Gateway), and digitized with a Digidata 1332A (Axon Instruments). Recording solutions contained 90 mM NaCl, 2 mM MgCl2 and 10 mM HEPES (pH 7.4), adjusted with NaOH. Data were analyzed using Clampfit (Axon Instruments).
SAXS data were collected at ALS beamline 12.3.1 (Lawrence Berkeley National Laboratory) on protein complexes at 8 mg ml–1 and 16 mg ml–1. Details of data collection and data processing can be found in Supplementary Methods.
Supplementary Figure 1 (a) Monomer of the KChIP1/T1N fusion23. KChIP1 amino acids are in slate blue with the exception of the H10 helix which is green. T1N (residues 1-21) (light orange) and N-terminal residue T1N are indicated. Calcium ions are shown as magenta spheres. (b) Ribbon diagram of a superposition of the Frequenin EF2 (dark blue), which is calcium bound, and KChIP1 EF2 in the KChIP1/Kv4.3 T1 domain complex (cyan). T1N (orange) is shown. Frequenin side chains that participate in calcium co-ordination are shown in stick representation. (c) Frequenin structure27 (skyblue). Three polyethylene glycol molecules bound in the hydrophobic groove are shown as sticks with carbon (skyblue) and oxygen (red) atoms indicated. Calcium ions are shown as magenta spheres.
Supplementary Figure 2 T1N helix and linker structure. (a) 2Fo-Fc omit map (green) contoured at 1.0 σ showing the density for the T1N aromatic residues (orange) and KChIP1 (cyan). Labels indicate select residues. Similar quality density was seen for Trp19. (b) NCS averaged 2Fo-Fc omit map density (green) derived for the T1N linker contoured at 1.0 σ from a 4.2 Å dataset. This dataset was used as the map shows better density for this region. (c) Selenomethionine anomalous signal shown from NCS averaged anomalous difference maps at 4.2 Å. Maps are contoured at 3.5 σ (pink), 4.0 σ (hot pink), and 4.5 σ (black). Kv4.3 and KChIP1 elements are labeled and colored as before. Methionine side chains are shown as sticks.
Supplementary Figure 3 Solution analysis of the KChIP1/Kv4.3 T1 complex. (a) SAXS data (top) and distribution function plot (bottom) for the complex. s = 4πsin(θ)/λ, where λ is the X-ray wavelength. (b) Top two panels, ab initio model of the KChIP1/Kv4.3 T1 complex (spheres). Bottom panel, comparison of the data and the calculated scattering from the ab initio model (χ2 = 3.46). (c), Dynamic light scattering data for the KChIP1/Kv4.3 T1 domain complex (16 mg ml-1 in a buffer of 50 mM KCl, 1 mM CaCl2, 50 μM ZnCl2, 30 mM DTT, 10 mM Na+-HEPES, pH 7.4) shows a single, monodisperse species in solution. The measured molecular weight, 148 kDa is in excellent agreement with the expected molecular weight of a (KChIP1)4(Kv4.3 T1 domain)4 complex (151 kDa). (d, e, and f) Comparison of the SAXS data with the calculated profiles for d, the T1 tetramer alone (χ2= 12.41), e, KChIP1 alone (χ2= 10.32), and f, the hexadecamer seen in the asymmetric unit of the KChIP1/Kv4.3 T1 domain crystal structure (χ2= 9.15).
Supplementary Figure 4 (a) Superdex 200 gel filtration elution profiles for purified KChIP1 37-216 and EF mutants. Molecular weights were determined using six proteins of known mass as standards. EF2, EF4 and EF2/EF4 mutants behave similar to wild-type (blue line). EF3 behaves similar to the EF2/EF3 mutant (green line). EF3/EF4 behaves similarly to EF2/EF3/EF4 (red line). Expected molecular weights are: KChIP1 27-216, 21 kDa; DnaK 70 kDa. (b) Top, 15% SDS-PAGE of KChIP1, EF3/EF4, EF2/EF3/EF4, and wild-type peak fractions from gel filtration. Bottom, Western-blot of the top panel using an anti-DnaK antibody. (c) Dynamic Light Scattering analysis of wild-type KChIP1 37-216 in the presence and absence of calcium.
Protein expression and purification
Kv4.3 1-143/KChIP1 37-216 complex
Rat Kv4.3 1-1431 was cloned into a modified pET28 vector containing a His6 tag, maltose binding protein, and a TEV cleavage site on the N-terminus of the protein sequence (pET28-HMT)2. hKChIP1 (residues 37-216) (IMAGE clone) was cloned into pEGST vector lacking the GST tag3. Vectors were co-transformed in BL21(DE3)pLysS E. coli, grown in 2YT media at 37° C, and induced (0.6-1.0 OD600 nm) with 0.4 mM IPTG for 4 hrs. Cells were resuspended in lysis buffer (150 mM KCl, 10% w/v sucrose, 25 mM n-octyl-β-D-glucopyranoside, 20 ng/μl lysozyme, 25 ng/μl Dnase I, 5 mM MgCl2, 1 mM CaCl2, 50 μM ZnCl2, 1 mM PMSF, 100 mM Tris-HCl, pH 8.8), disrupted by sonication, and centrifuged (30’, 31000 g, 4 °C, Beckman JA 25.50). Protamine sulfate at a final concentration of 2 mg/ml was added to the supernatant, incubated on ice for 10’, centrifuged (15’, 12000g, 4 °C, Beckman JA 12), and the supernatant loaded into a Poros20MC Ni2+ column (Applied Biosystem) equilibrated with buffer A (250 mM KCl, 1mM CaCl2, 50 μM ZnCl2, 10 mM HEPES-NaOH, pH 7.4). Protein was eluted in buffer A supplemented with 300 mM imidazole using a linear gradient over one column volume. 10 mM β-mercaptoethanol was added to the protein solution and the affinity tag was removed with TEV protease4 (overnight, room temperature). Cleavage reaction was dialyzed against buffer B (50 mM KCl, 1 mM CaCl2, 50 μM ZnCl2, 10 mM β-mercaptoethanol, 10 mM TRIS-HCl, pH 8.8), loaded onto a HiLoad Q 16/10 column (GE Healthcare), and eluted in a linear gradient from 50 mM -0.5 M KCl over twelve column volumes. The complex was further purified on an amylose column (New England Biolabs) in buffer A supplemented with 10 mM β-mercaptoethanol to remove traces of MBP and followed by gel filtration on a Superdex 200 (GE Healthcare) in buffer A + 10 mM β-mercaptoethanol. Protein was concentrated to ~15 mg/ml and stored at 4° C in 50 mM KCl, 1 mM CaCl2, 50 μM ZnCl2, 30 mM DTT, 10 mM HEPES-NaOH, pH 7.4. Point mutations were made by QuikChange (Stratagene) and expression and purification were performed as described for the wild-type complex.
Selenium methionine complex
Selenium methionine substituted complex was expressed in BL21(DE3)pLysS using autoinduction media5 (40 hrs., 37°C). Purification was performed as described for the native complex except that all buffers contained 5 mM methionine to prevent selenium oxidation.
hKChIP1 37-216 and EF mutants were cloned into pET28-HMT. Expression and purification were performed as described for the complex with the difference that buffers did not contain Zn2+ and β-mercaptoethanol. Proteins were stored at 4° C in 250 mM KCl, 1 mM CaCl2, and 10 mM HEPES-NaOH, pH = 7.4.
rKv4.3 1-143 and hKChIP1 (1-216) or EF mutants were co-expressed in E.coli and lysates were prepared as described above. Cell lysates were loaded on Amylose (New England Biolabs) resins equilibrated with buffer A (250 mM KCl, 1 mM KCl, 50 μM ZnCl2, and 10 mM HEPES-NaOH, pH 7.4). The complex was eluted in buffer A supplemented with 10 mM maltose and samples were loaded on 15% SDS-PAGE and stained with comassie brilliant blue R250.
Crystallization and Structure Determination
Prior to crystallization samples were centrifuged (30’, 70000g, 4°C, Beckman TLA 120.2). Dynamic light scattering (Dynapro, Wyatt Technologies) indicated that samples used for crystallization had <15% polydispersion. Crystals of the Kv4.3 1-143/KChIP1 37-216 K160A/K167A and selenomethionine derivative complexes were obtained at room temperature by hanging drop vapor diffusion using 1 μl of protein solution and 1 μl of 10-18% PEG 3000, 200 mM NaCl, 100 mM Bis-Tris, pH 6.5. Crystals used for data collection appeared in 7-10 days. Crystals were flash frozen in mother liquor supplemented with 20% glycerol. Data were collected at <100K at the Advanced Light Source (ALS) beamline 8.3.1 at the Lawrence Berkeley National Laboratories (LBNL). Diffraction images were indexed using MOSFLM6 and DENZO7 and scaled using SCALA8. Initial phases were obtained by using molecular replacement MOLREP9 with the Kv4.3 T1 domain tetramer (1S1G). Data collection wavelengths were 1.116 Å for native data and 0.9796 Å for SeMet data. Manual rebuilding correctly identified the position of the KChIP H2 helix. At this stage, we used the KChIP1 monomer (1S6C) with residues 159-189 deleted for molecular replacement. Refinement was carried out by iterative cycles of manual building in O910 and restrained refinement in Refmac58. Residues 3-20 and 39-140 of the T1 domain and residues 38-158 and 171-211 were refined using tight NCS restraints; the remaining loops were refined with medium restraints. Omit maps were generated using CNS11. Figures were displayed and rendered with Pymol12. The Ramachandran statistics for the structure are as follows: Most favored regions, 82.4 %; Additional allowed regions, 16.1%; Generously allowed regions, 1.5%, Disallowed regions, 0.0 %.
Circular Dichroism and Dynamic Light Scattering analysis
For the CD and DLS experiments, protein concentrations were determined by absorbance13.
Small angle X-ray scattering (SAXS)
Small-Angle X-Ray Scattering (SAXS) data were collected at ALS beamline 12.3.1 LBNL Berkeley, California. Complexes were 8 mg ml-1 and 16 mg ml-1 in a buffer containing 50 mM KCl, 1 mM CaCl2, 50 μM ZnCl2, 30 mM DTT, 10 mM Na+-HEPES, pH 7.4. Buffer scattering was subtracted and low- and high-angle curves were merged over the s range of 0.01-0.3 Å-1 using PRIMUS14. Radius of gyration (Rg) was evaluated using the Guinier approximation and from the entire scattering curve with GNOM15. GNOM also provides the distribution function p(r). To determine Dmax, the p(r) function was computed while constraining the function to go to zero at rmax, where rmax was varied from 100 to 200 Å in 5 Å increments. The rmax that yielded the highest “total estimate” value in combination with a plausible p(r) function was taken as Dmax. Ab initio models were generated using GASBOR16. Ten independent runs were performed imposing P1, P2 and P4 symmetry. Models were aligned and averaged using DAMAVER17. Crystal structures of Kv4.3 29-143 (1S1G) and KChIP1 37-216 (1S1E) were manually fit into the model using XtalView18. Theoretical scattering curves and fits to the experimental scattering curve from atomic coordinates were calculated using CRYSOL19.
Wild-type KChIP1 and EF3/EF4 and EF2/EF3/EF4 mutants were run on a 15% SDS-PAGE and afterwards transferred to a nitrocellulose membrane. Samples were probed with anti-DnaK antibodies (Stressgen biotechnologies) and detected with ECL (GE Healthcare).
1. Tsaur, M.L., Chou, C.C., Shih, Y.H. & Wang, H.L. Cloning, expression and CNS distribution of Kv4.3, an A-type K+ channel alpha subunit. FEBS Lett 400, 215-20 (1997).
2. Van Petegem, F., Clark, K.A., Chatelain, F.C. & Minor, D.L., Jr. Structure of a complex between a voltage-gated calcium channel beta-subunit and an alpha-subunit domain. Nature 429, 671-5 (2004).
3. Kholod, N. & Mustelin, T. Novel vectors for co-expression of two proteins in E. coli. Biotechniques 31, 322-3, 326-8. (2001).
4. Kapust, R.B. et al. Tobacco etch virus protease: mechanism of autolysis and rational design of stable mutants with wild-type catalytic proficiency. Protein Eng 14, 993-1000. (2001).
5. Studier, F.W. Protein production by auto-induction in high density shaking cultures. Protein Expr Purif 41, 207-34 (2005).
6. Leslie, A.G.W. Recent changes to the MOSFLM package for processing film and image plate data. Joint CCP4 + ESF-EAMCB Newsletter of Protein Crystallography 26(1992).
7. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Meth. Enzy. 276, 307-326 (1997).
8. Collaborative Computational Project, N. The CCP4 suite: Programs for protein crystallography. Acta Crystallogr D Biol Crystallogr 50, 760-763. (1994).
9. Vagin, A. & Teplyakov, A. MOLREP: an automated program for molecular replacement. J. Appl. Cryst. 30, 1022-1025 (1997).
10. Kleywegt, G.J. & Jones, T.A. Efficient rebuilding of protein structures. Acta Crystallogr D Biol Crystallogr 52, 829-32 (1996).
11. Brunger, A.T. et al. Crystallography & NMR system: A new software suite for macromolecular structure determination. Acta Crystallogr D Biol Crystallogr 54, 905-21. (1998).
12. DeLano, W.L. The PyMOL Molecular Graphics System. (DeLano Scientific, San Carlos, CA, 2002).
13. Edelhoch, H. Spectroscopic determination of tryptophan and tyrosine in proteins. Biochemistry 6, 1948-54. (1967).
14. Konarev, P.V., Volkov, V.V., Sokolova, A.V., Koch, M.H.J. & Svergun, D.I. PRIMUS: a Windows PC-based system for small-angle scattering data analysis. J. Appl. Cryst. 36, 1277-1282 (2003).
15. Svergun, D.I. Determination of the regularization parameter in indirect-transform methods using perceptual criteria. J. Appl. Cryst. 25, 495-503 (1992).
16. Svergun, D.I., Petoukhov, M.V. & Koch, M.H. Determination of domain structure of proteins from X-ray solution scattering. Biophys J 80, 2946-53 (2001).
17. Volkov, V.V. & Svergun, D.I. Uniqueness of ab initio shape determination in small-angle scattering. J. Appl. Cryst. 36, 860-864 (2003).
18. McRee, D.E. XtalView/Xfit--A versatile program for manipulating atomic coordinates and electron density. J Struct Biol 125, 156-65. (1999).
19. Svergun, D.I., Barberato, C. & Koch, M.H.J. CRYSOL - a program to evaluate X-ray solution scattering of biological macromolecules from atomic coordinates. J. Appl. Cryst. 28, 768-773 (1995).
We thank D. Julius and F. Van Petegem for comments on the manuscript, J. Holton at ALS beamline 8.3.1 for data collection help, L.Y. Jan (University of California San Francisco) for the Kv4.3 clone and Minor laboratory members for support. This work was supported by awards to D.L.M. from the McKnight Foundation for Neuroscience, Rita Allen Foundation, Searle Scholars Award, Arnold and Mabel Beckman Foundation and US National Institutes of Health. G.L.H. was supported by US National Cancer Institute grant CA92584 and US Department of Energy contract DE-AC03-76SF00098. D.L.M. is a McKnight Scholar, Rita Allen Scholar, Searle Scholar and Beckman Young Investigator.
Accession codes. Protein Data Bank: Coordinates and structure factors have been deposited with accession code 2I2R.
Note: Supplementary information is available on the Nature Structural & Molecular Biology website.
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial interests.