Ito Channels Are Octomeric Complexes with Four Subunits of Each Kv4.2 and K+ Channel-interacting Protein 2*

Mammalian voltage-gated K+ channels are assemblies of pore-forming α-subunits and modulating β-subunits. To operate correctly, Kv4 α-subunits in the heart and central nervous system require recently identified β-subunits of the neuronal calcium sensing protein family called K+ channel-interacting proteins (KChIPs). Here, Kv4.2·KChIP2 channels are purified, integrity of isolated complexes confirmed, molar ratio of the subunits determined, and subunit valence established. A complex has 4 subunits of each type, a stoichiometry expected for other channels employing neuronal calcium sensing β-subunits.

KChIPs (and frequenin) (16,17) have been shown to assemble with Kv4 subunits in stable fashion leading to increased current density due to enhanced surface expression, activation at more hyperpolarized potentials, slowed inactivation, and speeded recovery from inactivation (1,2,16). These ␤-subunits permit kinase regulation (18), control trafficking, and alter surface half-life (6). So, the increase in I to across canine cardiac ventricular wall (from endo-to epicardium) reflects a parallel gradient in KChIP2 expression (3), and transgenic mice devoid of KChIP2 have no measurable I to , a prolonged action potential, and susceptibility to ventricular arrhythmia (4).
Functional and structural studies of ␣-subunits demonstrate that K ϩ channel conduction pores are formed by assembly of 4 pore loops about a central axis (19). Conversely, K ϩ channel accessory subunits vary in their nature (soluble, peripheral, or membrane-spanning) and number per complex (20). To explore the structural basis for KChIP function, the stoichiometry of Kv4.2⅐KChIP2 channels was determined using a stepwise strategy: significant amounts of intact, functional Kv4.2⅐ KChIP2 complexes were produced and isolated; functional and structural integrity of the purified material was confirmed by a toxin binding assay; and, thereafter, subunit valence was established by gradient centrifugation and direct amino acid analysis.

MATERIALS AND METHODS
Molecular Biology-Human Kv4.2 and KChIP2 were expressed in pRAT, a dual purpose vector with a cytomegalovirus promoter for mammalian expression and a T7 promoter for cRNA synthesis. Charybdotoxin (CTX)-sensitive Kv4.2 (Kv4.2*) was produced by plaque-forming unit-based mutation (Stratagene, La Jolla, CA); the pore segment containing mutations was isolated with NruI and BamHI and replaced into a wild-type human Kv4.2 backbone. The 1D4 epitope was introduced by creating an Mlu1 site in place of the termination codon and inserting a nucleotide compatible to Mlu1 and BglII encoding a linker and the epitope (RVPDGDPDETSQVAPAX). cRNA was synthesized after linearization with NotI using an mMessage mMachine kit (Ambion, Austin, TX). cRNAs were quantified by spectroscopy and gel electrophoresis. The gene variants used in this study were human Kv4.2 (accession number AH009258), KChIP2 (accession number AF199598), and Kv1.3 (accession number NM_002232).
Electrophysiology and Data Analysis-Oocytes from Xenopus laevis frogs were defolliculated by collagenase and injected with 0.2 ng of Kv4.2 or Kv4.2* and 0.4 ng of KChIP2 cRNA. For CTX block studies less cRNA was injected for mixed complexes so macroscopic current was like that for ␣-subunits alone. Currents were studied at 12-24 h. Whole-cell currents were measured by two electrode voltage clamp (Geneclamp 500, Axon Instruments, Union City, CA). Electrodes were filled with 3 M KCl and had resistance of 0.1-1.2 M⍀. Data were sampled at 1 kHz. Data recording was performed using Clampex v8.0 and assessed with Clampfit v8.0, Excel, and Origin 6.0. All experiments were performed at room temperature. Standard bath solution was ND-96: 96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM HEPES/NaOH, pH 7.5) and supplemented with 0.1% bovine serum albumin for CTX studies. CTX binding kinetics were examined with voltage pulses to 40 mV for 500 ms from a resting membrane voltage of Ϫ80 mV repeated every 20 s. Current-voltage curves were evoked by depolarizing from a holding potential of Ϫ100 mV to test potentials of Ϫ80 to 70 mV with 10 mV steps lasting 500 ms every 5 s. Steady state inactivation was examined from a holding potential of Ϫ100 mV with test pulses from Ϫ120 to 10 mV held for 2.5 s with a second pulse at 40 mV to measure currents that were not inactivated. Recovery from inactivation was measured by driving channels to an inactivated state at 40 mV, hyperpolarizing to Ϫ100 mV, and then applying a second pulse to 40 mV for various intervals (increments of 5 and 50 ms). Inactivation was fit to a double exponential equation. Recovery from inactivation was fit to a single exponential. Activation V 0.5 and V s were calculated by fitting the conductance-voltage relationship to a Boltzmann function. Similarly, inactivation V 0.5 and V s were calculated by fitting the normalized currentvoltage relationship to a Boltzmann function.
Whole-cell CTX Binding-3 H-CTX binding to cells was performed as described previously (22). Briefly, cell samples were divided into control and test groups. Control assessed nonspecific binding of 3 H-CTX (36 nM) that took place in the presence of 1 M non-radioactive CTX. Test groups measured total 3 H-CTX binding and specific binding was determined by subtracting the nonspecific from the total binding. All steps were performed at 4°C. Cells were exposed to 3 H-CTX in binding buffer (20 mM KCl, 10 mM NaPi, 100 mM sucrose, and 1 mg/ml bovine serum albumin) for 30 min, collected by centrifugation for 1 min at 3,000 rpm, and the supernatant was removed. The cell pellet was washed in 100 l binding buffer, centrifuged, and after removal of supernatant, suspended in 100 l of buffer, transferred to a scintillation vial with 10 ml of scintillation fluid (SafeScint, American Bioanalytical, Natick, MA) for liquid scintillation counting.
CTX Binding to Purified Complexes-A 50 l elution of purified material was exposed to 36 nM 3 H-CTX with or without 1 M unlabelled CTX. As above, the two samples were used to assess nonspecific and total binding of 3 H-CTX. Binding was performed at 4°C for 30 min, 8 h, 16 h, or 24 h, after which unbound toxin was removed with an Amicon Centricon 100 (Millipore, Billerica, MA) by centrifugation for 10 s at 9,000 rpm and a wash with 20 l wash buffer with only 100 mM NaCl by centrifugation. The filter was then placed in a liquid scintillation vial and studied in a scintillation counter as above. The dosetitration assay was performed in a similar manner with 25 l aliquots of purified channel complex. 0, 0.45, 1.33, 4, 12, or 36 nM 3 H-CTX was applied to a control and test sample for 30 min at 4°C and processed as above.
Amino Acid Analysis-After SDS-PAGE and staining with Coomas-  sie Brilliant Blue, bands corresponding to Kv4.2* and KChIP2 and a blank portion of the gel were excised with a sterile scalpel. Samples were hydrolyzed in 6 N HCl and 0.2% phenol for 16 h at 115°C with 2 nmol norleucine/100 l HCl as an internal standard. After hydrolysis, amino acids were quantified on a Beckman 7300 Amino Acid Analyzer via ion-exchange and post-column ninhydrin detection (Yale University Keck Facility). After quantification, some amino acids are not analyzed due to intrinsic limitations of the method: these include His, which co-elutes with Tris in the SDS-PAGE buffer, Gly, which is in SDS-PAGE buffer, and Arg, which is variably obscured by ammonia in gel samples. Amino acids residues determined for the blank gel sample were subtracted from subunit samples to correct for background signal. 9 -12 amino acids for each subunit were observed to be within 10% of predicted value in each trial and were studied: Asn, Asp, Thr, Gln, Glu, Pro, Ala, Val, Ile, Leu, Phe, and Lys.

RESULTS
First, a variant of human Kv4.2 ␣-subunits (Kv4.2*) was produced that maintained the functional attributes of wildtype but had two modifications, a pore binding site for the peptide toxin CTX and an 8 residue C-terminal epitope tag (1D4, ETSQVADA). The rationale for these embellishments was as follows. A single CTX molecule binds in the ion conduction pore of toxin-sensitive K ϩ channels to occlude the permeation pathway only when 4 permissive ␣-subunits are assembled in correct fashion; as a result, small changes in pore (or toxin) structure are registered with great sensitivity as altered blockade (23-25). Thus, CTX and related toxins were used to locate the external aspect of the conduction pore (26), assess its dimensions (27,28), count channels (22), and, as here, assess the integrity of purified channel protein (19). The 1D4 epitope allows purification of membrane proteins under mild conditions via monoclonal antibody binding and peptide elution (21).
The CTX-sensitive variant of Kv4.2 was produced by introduction of 3 mutations in the pore loop (Fig. 1a) chosen in an iterative fashion based on residues found to be important for high-affinity toxin binding in other K ϩ channels (29, 30). Although wild-type Kv4.2 channels are insensitive to application of CTX, those with Kv4.2* ␣-subunits show half-maximal block (K i ) at 0.35 Ϯ 0.04 nM (Fig. 1, b and c). Similarly, wild-type Kv4.2⅐KChIP2 channels are insensitive to toxin whereas those with Kv4.2* ␣-subunits showed K i ϭ 0.76 Ϯ 0.03 nM (Fig. 1, b  and c). As CTX is external and KChIP2 subunits are intracellular, the increase in K i observed on KChIP2 incorporation was unexpected. Relaxation to equilibrium blockade upon acute toxin application revealed KChIP2 to have little effect on the toxin on-rate (K on ϭ 18.1 Ϯ 1.4 M Ϫ1 s Ϫ1 versus K on ϭ 24.5 Ϯ 2.3 M Ϫ1 s Ϫ1 with KChIP2, n ϭ 8 -10) but to destabilize bound toxin, increasing its off-rate ϳ3-fold (K off ϭ 8.0 Ϯ 0.6 ϫ 10 Ϫ3 s Ϫ1 versus K off ϭ 22.9 Ϯ 2.6 ϫ 10 Ϫ3 s Ϫ1 ). This indicates small KChIP2-induced changes in the external pore of the closed channel (the predominant state in this protocol).
Despite modifications to confer CTX sensitivity and append the epitope tag, channels with Kv4.2* behaved almost like those with wild-type subunits (Table I and Fig. 2). Kv4.2* channels are like wild-type in their voltage-dependence for half-maximal activation and half-maximal steady state inactivation. Although small differences in inactivation kinetics were observed for slow ( slow ) and fast ( fast ) time constants, these were without associated changes in their relative ratio, or rates of recovery from inactivation. Moreover, the archetypal effects of KChIP2 were unaltered by the ␣-subunit mutations. Thus, KChIP2 produced similar increases in current density and rates of recovery from inactivation with both wild type Kv4.2 and Kv4.2*, and similar decreases in rates of inactivation (due to increased contribution of the slow component) with expected small shifts in the voltage required to achieve half-maximal activation and half-maximal steady state inactivation.
Radioactive CTX ( 3 H-CTX) was synthesized as previously reported (22) and used first to monitor and optimize surface expression of channels on COS7 cells (see "Materials and Methods"). As expected based on measurement of channel currents, KChIP2 increased the absolute number of channels on the cells ϳ2.6-fold, from 1.6 Ϯ 0.2 pmol/plate (n ϭ 4) to 4.0 Ϯ 0.2 pmol/plate (n ϭ 11). Purification of Kv4.2*⅐KChIP2 complexes was achieved via the 1D4 epitope on Kv4.2* with anti-1D4 antibodies covalently bound to beads (Fig. 3). Thus, COS7 cells induced to transiently express Kv4.2* and KChIP2 at high levels were solubilized with the detergent CHAPS (2.5%), and complexes isolated by binding, washing with buffer, and elution with 1D4 peptide (see "Materials and Methods"). This led to co-purification of Kv4.2* and KChIP2 (Fig. 3, a-c). Western blot analysis revealed excess KChIP2 in soluble material (SM) and unbound flow though (FT) whereas Kv4.2*⅐KChIP2 channels were enriched in fractions collected by peptide elution (E) (Fig. 3a). Staining with Coomassie Blue confirmed the E2 fraction to be rich in the two channel subunits (Fig. 3b).
Purified material was subjected to centrifugation through a glycerol gradient and studied by Western blot analysis after SDS-PAGE to estimate the size of Kv4.2*⅐KChIP2 channels. The complex had a molecular mass between 232-440 kDa in trials with four separate preparations (Fig. 3c). Because Kv4.2* subunits carry no carbohydrate, each 642 residue protein has a predicted mass of ϳ72 kDa and each tetramer a mass of ϳ288 kDa (an underestimate due to detergent binding (21)). Because KChIP2 has 252 residues and a predicted mass of ϳ28 kDa this result argues that complexes contain between 1 and 6 KChIP2 subunits (ϳ316 -456 kDa). The predicted mass for an octameric complex with 4 Kv4.2* subunits and 4 KChIP2 subunits is 400 kDa.
To verify the stability and integrity of purified Kv4.2*⅐ KChIP2 channels over time, binding of 3 H-CTX to isolated complexes was assessed. After 24 h at 4°C, ϳ90% of the binding activity was retained (Fig. 4a). Indeed, despite residence in detergent rather than a lipid membrane, purified complexes bound 3 H-CTX with high affinity. Binding studies with four separate preparations Kv4.2*⅐KChIP2 were well fit to a singlesite binding (Langmuir) isotherm with a K d ϳ4.8 nM (Fig. 4b).
Because CTX binding occurs with high affinity only when pores are properly assembled, the results provide direct evidence that the isolated complexes are a good approximation of the channels in plasma membranes and indicate their homogeneity because a single high affinity toxin site is apparent.
To accurately assess the subunit composition of purified Kv4.2*⅐KChIP2 complexes, the two subunits were separated by SDS-PAGE, identified by staining with Coomassie Brilliant  (Table II). The band with a star is a minor contaminant (Ͻ1%) of 1D4 purifications (21). c, E2 fraction applied to a 10 -40% glycerol gradient subjected to centrifugation and examined by 10% SDS-PAGE and Western blot analysis with anti-1D4 and anti-KChIP2 antibodies. Peak fractions for Kv4.2*⅐KChIP2 complexes are between markers at 232 and 440 kDa (catalase and ferritin, respectively). Blue dye, and the bands excised from the gel for amino acid analysis by hydrolysis, ion-exchange chromatography, and ninhydrin detection (Table II). Because the amino acid sequences of Kv4.2* and KChIP2 were known, the expected and observed amino acid content of the bands could be compared and the moles of each subunit determined independently from 9 -12 residues in each of 3 independent trials. The molar ratio of Kv4.2*:KChIP2 in purified complexes was 1:1 (Table II); similar results are noted in a companion report for channels with a KChIP2 variant and others with a Kv4.2* mutant. 2 Purified channels are shown to have 4 ␣-subunits because CTX binds with high affinity (Fig. 4). A 1:1 molar ratio therefore indicates that Kv4.2*⅐KChIP2 complexes contain 4 KChIP2 subunits.

DISCUSSION
Accessory ␤-subunits are a fundamental feature of K ϩ channels that determine channel location, abundance, sensitivity to stimulation, and pharmacology in vivo (6,20,32). Here we demonstrate that regulatory KChIP2 subunits assemble with pore-forming Kv4.2 subunits in 4:4 complexes when over-expressed in tissue culture cells to produce voltage-gated K ϩ channels like those in native cells (such as cardiac I to (3,4)). A 4:4 subunit arrangement appears to be the natural and stable valence for Kv4.2*⅐KChIP2 channels first because it was found with four separate large-scale preparations and second because KChIPs could not be separated from ␣-subunits without complete dissociation of the complexes to monomer form (not shown). KChIP2 increases trafficking to the surface, channel half-life, and detergent solubility of mixed complexes formed with Kv4.2 (6). This suggests why we could purify significant amounts of Kv4.2*⅐KChIP2 channels but did not succeed when Kv4.2* ␣-subunits were expressed alone (not shown). We expect KChIPs and related neuronal calcium sensing proteins (such as frequenin (16)) to assemble with other K ϩ channel ␣-subunits with 4:4 stoichiometry to yield similar effects on stability, structure, and function.
In a companion study, 2 Kv4.2*⅐KChIP2 complexes purified by the strategy described here were found amenable to visualization by negative stain electron microscopy; demonstration here that isolated complexes maintain structural integrity and have 4:4 subunit valence allowed three-dimensional images of the channels to be reconstructed with a resolution of 21 Å. KChIP2 incorporation was found to create an ϳ36 ϫ 115 ϫ 115 Å, intracellular fenestrated rotunda: 4 peripheral columns that extend down from the membrane-embedded portion of the channel to enclose the central Kv4.2 "hanging gondola" (a platform held beneath the transmembrane conduction pore by 4 internal columns) (21,31,42,43). To reach the pore from the cytosol, ions pass through one of four external fenestrae to enter the rotundal vestibule and then cross one of four internal windows in the gondola. The location of KChIP2 subunits in the structure is lateral to the gondola platform and does not overlap with the sub-platform locale for the Kv␤2 subunits apparent in crystal structures (33). It is as yet unknown whether a single channel complex can carry more than one type of accessory subunit and if subunit stoichiometry is thereby varied. These issues are under study by the methods described here.