Purification, visualization, and biophysical characterization of Kv1.3 tetramers.

The voltage-gated K+ channel of T-lymphocytes, Kv1.3, was heterologously expressed in African Green Monkey kidney cells (CV-1) using a vaccinia virus/T7 hybrid expression system; each infected cell exhibited 104 to 5 × 105 functional channels on the cell surface. The protein, solubilized with detergent (3-[cholamidopropyl)dimethylammonio]-1-propanesulfonic acid or cholate), was purified to near-homogeneity by a single nickel-chelate chromatography step. The Kv1.3 protein expressed in vaccinia virus-infected cells and its purified counterpart are both modified by a ∼2-kDa core-sugar moiety, most likely at a conserved N-glycosylation site in the external S1-S2 loop; absence of the sugar does not alter the biophysical properties of the channel nor does it affect expression levels. Purified Kv1.3 has an estimated size of ∼64 kDa in denaturing SDS-polyacrylamide electrophoresis gels, consistent with its predicted size based on the amino acid sequence. By sucrose gradient sedimentation, purified Kv1.3 is seen primarily as a single peak with an approximate mass of 270 kDa, compatible with its being a homotetrameric complex of the ∼64-kDa subunits. When reconstituted in the presence of lipid and visualized by negative-staining electron microscopy, the purified Kv1.3 protein forms small crystalline domains consisting of tetramers with dimensions of ∼65 × 65 Å. The center of each tetramer contains a stained depression which may represent the ion conduction pathway. Functional reconstitution of the Kv1.3 protein into lipid bilayers produces voltage-dependent K+-selective currents that can be blocked by two high affinity peptide antagonists of Kv1.3, margatoxin and stichodactylatoxin.

Voltage-gated K ϩ (Kv) 1 channels regulate membrane potential and thereby control many biological processes in many cell types from bacteria to humans (reviewed in Refs. 1, 2). Four families of mammalian genes, Kv1-Kv4, comprising 19 members, encode a diversity of Kv channels (2). Each of these proteins contains 500 -600 amino acids, typically long N and C termini, six putative transmembrane segments (S1-S6), and an additional membrane-associated loop (P-region) between S5 and S6. Several functional domains have been identified by site-directed mutagenesis in these Kv proteins (see Ref. 2), and the outer vestibule of the pore has been topologically mapped using structurally defined high affinity scorpion toxins as molecular calipers (3)(4)(5)(6).
Previous attempts to overexpress and purify Kv protein for biochemical and biophysical studies have met with limited success. The Drosophila Shaker protein has been transiently expressed at high levels in COS cells, but the poor transfection efficiency (10% of cells express protein) precludes its use as a reliable source for protein purification (7). Although expression with the baculovirus system has been more efficient, yields from this system have still been insufficient (8 -10). In addition, the quality of the overexpressed protein is seriously compromised since a significant portion of this protein is localized within inclusion bodies, rendering it insoluble in nondenaturing detergent, and very little of the protein is glycosylated (7,9,10). Lipid-bilayer reconstitution of these purified channels for biophysical characterization has not been demonstrated.
The goals of this study are as follows: (a) to characterize a mammalian heterologous overexpression system for the purification of appropriately glycosylated, membrane-associated mammalian Kv1 protein; (b) to determine the biophysical properties of purified Kv1 protein reconstituted into lipid bilayers; and (c) to verify the tetrameric nature of the channel using sucrose density sedimentation and negative-staining electron microscopy. The Kv protein we have used for this analysis is Kv1.3, which represents the type "n" channel of T-lymphocytes, and regulates the membrane potential of these cells (11)(12)(13)(14)(15)(16). Blockers of Kv1.3, which include the high affinity peptide toxins margatoxin (MgTX) and stichodactylatoxin (ShK), suppress T-cell activation, making this channel an excellent therapeutic target for novel immunosuppressive agents (12)(13)(14)(15)(16).
In this study we have exploited a heterologous vaccinia virus (VV)-based expression system to overexpress Kv1.3 protein in mammalian cells, where its posttranslational processing is likely to resemble its native counterpart in T-cells. Utilizing this expression system, recombinant Kv1.3 protein was purified to near-homogeneity from a membrane fraction, and this protein appeared biochemically uniform, bearing a core glyco-sylation moiety and associating as an apparent tetramer. Additionally, this protein produced voltage-dependent, K ϩ -selective, MgTX-and ShK-sensitive currents when reconstituted into lipid bilayers and formed small crystalline domains, visible by electron microscopy, composed of 65 ϫ 65-Å tetramers.
Antibodies-A horseradish peroxidase-labeled mouse monoclonal antibody (Ab) specific for a 12-residue epitope derived from gene 10 of bacteriophage T 7 (T7-Tag, also referred to as anti-gene 10 Ab) was purchased from Novagen (Madison, WI), and Texas Red-conjugated donkey anti-mouse IgG was obtained from The Jackson Laboratories (Bar Harbor, ME).
Cell Lines and Vaccinia Viruses-African Green monkey kidney cells (CV-1), rat basophilic leukemic (RBL) cells, the WR strain of VV, and the recombinant VV which expresses T7 RNA polymerase under control of the p7.5 VV promoter (vTF7-3 or VV:T7; 17) were acquired from the ATCC (Rockville, MD). CV-1 cells were grown in Eagle's minimal medium with 10% fetal bovine serum, 2 mM L-glutamine, 10 units/ml penicillin G, and 10 g/ml streptomycin. RBL cells were grown in RPMI with 10% fetal bovine serum and 2 mM L-glutamine.

Vaccinia Virus for the Heterologous Expression of Kv1.3 Protein in Mammalian Cells
Generation of VV:Kv1.3 Construct and Recombination into Vaccinia-We generated the pTH1 vector by inserting a 144-base pair NcoI/HindIII fragment (the HindIII site was blunt-ended by a fill-in reaction) from the pTrcHis vector (a kind gift from Dr. Leonard Wittwer of the Invitrogen Corp., Sorrento Valley, CA) into the pTM1 vector (17,18) at NcoI/StuI sites. The 144-base pair insert contained sequences coding for an initiator methionine followed by a hexahistidine repeat, a serologically detectable epitope from bacteriophage T7 (gene 10), an enterokinase cleavage site, and a multiple cloning site. The coding region was flanked by a T7 promoter and terminator, and the 5Јnoncoding region was derived from encephalomyocarditis virus which provides an internal ribsome binding site for efficient translation initiation. The protein coding region of mouse Kv1.3 gene was inserted into pTH1 as a 2-kilobase BglII/EcoRI fragment excised from the pMK3T construct (11), and the integrity of the construct was confirmed by dideoxy sequencing. We recombined the Kv1.3-pTH1 plasmid into VV to generate the VV:Kv1.3 recombinant using standard methods (19).
Dual Infection of CV-1 Cells-CV-1 or RBL cells were doubly infected with 5-10 multiplicity of infection of VV:Kv1.3 and VV:T7 (see 19); in these cells, T7 RNA polymerase synthesis controlled by the VV early p7.5 promoter leads to T7-mediated transcription and Kv1.3 protein production. After 2-16 h, the cells were subjected to either patch clamp or immunofluorescence analysis or were harvested for protein purification.
Immunofluorescence Analysis-The anti-gene 10 Ab was used for cellular localization of Kv1.3. The specificity of this Ab was determined by Western blotting and peptide competition (data not shown), the expected size of the Kv1.3 protein being ϳ64 kDa. At selected times cells were pelleted, resuspended in phosphate-buffered saline, pH 7.4, with 4% paraformaldehyde, and placed on ice for 10 min. A firm cell pellet was made by spinning the cells at 10,000 ϫ g for 10 min. After an additional incubation on ice for 30 min, the supernatant was aspirated, and the cell pellet was gently removed from the microcentrifuge tube and infiltrated with 50% polyvinylpyrrolidone in 0.1 M phosphate buffer, pH 7.4, containing 2.3 M sucrose overnight at 4°C. Thick 0.5-m cryosections of the pellets were prepared (21) and then incubated with the anti-gene 10 Ab or control mouse IgG (1:100) for 1 h followed by incubation with 5 g/ml Texas Red-conjugated donkey anti-mouse IgG. Exposure-matched fluorescence micrographs were taken on a Nikon FXA microscope.
Nickel-Chelate Chromatography-All chromatography was performed using the SMART System (Pharmacia, Uppsala, Sweden). Chelating Sepharose was bound with nickel as recommended by the manufacturer and then washed with 10 column volumes of Native buffer. After loading the solubilized protein, the column was washed with Native buffer to an A 280 Ͻ0.02. The column was then washed with a High Salt buffer (1 M KCl, 50 mM imidazole, 2% CHAPS, 10 mM Tris, pH 7.4) to elute proteins which bind to the column through nonspecific ionic interactions, and washing was continued to an A 280 Ͻ0.02. Elution of Kv1.3 protein from the column was performed using a linear gradient of imidazole (50 -500 mM) in Native buffer. The protein peak containing Kv1.3 (at ϳ300 mM imidazole) was identified by dot blot chemiluminescence using the anti-gene 10 Ab.
SDS-PAGE, Protein Detection, and Quantitation-SDS-PAGE was performed according to the method of Laemmli (22). Samples were diluted in an equal volume of buffer (7.5% SDS, 10% glycerol, 50 mM Tris-HCl, pH 6.8, 0.1% bromphenol blue, 14.4 mM 2-mercaptoethanol) and loaded onto the gel without heating (to prevent the formation of large, insoluble aggregates). For immunoblots, proteins were transferred onto polyvinylidene difluoride membrane (Millipore, Bedford, MA) by electrotransfer in Towbin buffer (192 mM glycine, 25 mM Tris, 10% methanol). Immunoblots were performed by spotting 1-l samples onto nitrocellulose membrane (Schleicher & Schuell), enhanced chemiluminescence reagents from Amersham (Buckinghamshire, UK) being used for detection of the immune complexes. For quantitative immunoblotting, the scanned image was analyzed by NIH image (23). The positive control gene 10-extract (Novagen, Madison, WI) was used as a standard assuming 65% purity.
Silver staining of SDS-PAGE gels was performed by the diamine staining method (24). Quantification of protein was determined by A 280 , by the bicinchoninic assay (Pierce) and by quantitative chemiluminescence blotting (using anti-gene 10 Ab).

Glycosylation of Kv1.3
For [ 35 S]methionine labeling, dually infected CV-1 cells were incubated in methionine-deficient minimum Eagle's medium with 10% dialyzed fetal bovine serum, and 100 Ci of [ 35 S]methionine (1175 Ci/ mmol) in the presence or absence of 1 g/ml tunicamycin (C2 analog). After 12 h, cells were washed, harvested, and solubilized in RIPA buffer (150 mM NaCl, 1% Nonidet P-40, 0.1% SDS, 50 mM Tris, pH 8.0). To immunoprecipitate the Kv1.3 protein, cell nuclei were first pelleted and the supernatant precleared with 5 mg of protein-A Sepharose beads. One microliter of the anti-gene 10 Ab was then added and incubated for 1 h at ϩ4°C, and then 5 mg of protein A-Sepharose beads were added for another hour. The protein A-Sepharose was washed with RIPA buffer three times and resuspended in 30 l of sample buffer (see above). Samples were analyzed by fluorography following SDS-PAGE separation on a 10% polyacrylamide gel.
In vitro deglycosylation was performed using N-glycosidase F (Boehringer Mannheim). Samples were diluted in buffer containing 35 mM sodium phosphate, pH 7.0, and 5 mM EDTA, followed by the addition of SDS to a final concentration of 0.2%. After 1 h at room temperature, an equal volume of phosphate/EDTA buffer with 5% Nonidet P-40 was added along with 0.4 units of N-glycosidase F, and the mixture was incubated at 37°C overnight. Samples were separated by SDS-PAGE on a 10% polyacrylamide gel, transferred to polyvinylidene difluoride membrane, and immunoblotted as described above.

Determining the Multimeric Nature of Kv1.3
Sucrose Density-One-half milligram of each protein standard (apoferritin, alcohol dehydrogenase, bovine serum albumin, carbonic anhydrase) and ϳ10 g of purified Kv1.3 were layered on top of a 5-50% sucrose gradient (volume of ϳ12 ml) containing 20 mM Tris, pH 7.4, 50 mM KCl, 0.2 mM phenylmethylsulfonyl fluoride, and either 1.0% CHAPS or 0.6% cholate. Samples were centrifuged for 14 h at 220,000 ϫ g (20°C) and 10-drop fractions were subsequently collected. To determine which fractions contained Kv1.3, 10 l of each sample was filtered onto nitrocellulose using a slot blot apparatus and immunoblotted as described above. The scanned image of the negative was then analyzed with NIH Image (24) to quantitate the optical density of each spot. To visualize the protein standards, 15 l of each fraction was loaded with an equal volume of Laemmli buffer, run on a 10% SDS-polyacrylamide gel, and the gel then stained with Coomassie Blue.
The mass of Kv1.3 was estimated as described by Martin and Ames (25). Briefly, the sedimentation rates of proteins on continuous sucrose gradients are related to molecular masses by the equation S 1 /S 2 ϭ (MW 1 /MW 2 ) 2/3 , where S 1 and S 2 are the distances travelled by two proteins, and MW 1 and MW 2 are their molecular masses (25). We solved this equation for Kv1.3 relative to each protein standard and plotted each value against the molecular mass of the standard; the slope of the resulting line represents the estimated mass of Kv1.3.
Negative Staining EM of "Crystalline" Kv1.3 Protein-The purified Kv1.3 protein in CHAPS was reconstituted with dimyristoyl phosphatidylcholine at a lipid-to-protein ratio of 1:1 (w/w) and a protein concentration of 1 mg/ml, and the detergent was slowly removed by dialysis (26,27). For electron microscopy, a carbon-coated electron microscope grid was glow-discharged for 2 min prior to the application of 2 l of a Kv1.3 crystalline patch. The sample was kept on the grid for about 1 min before blotting with filter paper to remove excess buffer and left to air-dry. Samples were then coated with 1.5% uranyl acetate.

Biophysical Characterization of Lipid Bilayer Reconstituted and Purified Kv1.3 Protein
Purified Kv1.3 was reconstituted into small unilamellar vesicles using previously published methods (28). Briefly, PC/PE (2:1) solubilized in 60 mM CHAPS or cholate was added to 100-l aliquots of purified Kv1.3 to obtain a protein-to-lipid ratio of 1:1000, and then dialyzed overnight in a Slide-a-lyzer (Pierce) dialysis chamber against 3 liters of buffer containing 200 mM KCl and 25 mM HEPES, pH 7.4. Aliquots were stored at Ϫ20°C prior to use.
Bilayers were formed in Teflon chambers divided into two compartments, each having a volume of 2 ml. The compartments were separated by a 20 m-thick Teflon partition with a ϳ200-m hole pretreated with approximately 0.5 l of squalene (Atomergic Chemicals Corporation, Plainview, NY) (29,30). With the aqueous solutions below the hole, 10 l of 5 mg/ml phosphatidylethanolamine (Avanti Polar Lipids, Birmingham, AL), dried under argon, and dissolved in pentane (Aldrich) was spread over the aqueous phase. To form the bilayers, the solutions in the compartments were slowly raised above the level of the hole. Following bilayer formation (monitored by capacitance increase), an aliquot of reconstituted Kv1.3 protein was added to the high KCl (cis) side of the cell, and both chambers were stirred by magnetic fleas to promote vesicle interaction with the bilayer. The bilayer was voltage-clamped by an AXOPATCH 200 A controlled by an 80286 computer driving an Interactive Microware (State College, PA) ADALAB interface board controlled by locally written software.

RESULTS AND DISCUSSION
Biophysical Characterization of Kv1.3 Channels in VV-infected CV-1 Cells-We performed a detailed comparison of the functional properties of the Kv1.3-pTH1 channels expressed in CV-1 cells with channel "fingerprints" of the cloned Kv1.3 channel in mammalian cells (Table I) and Xenopus oocytes (11) and the native channel in mouse T-cells (31). These experiments were aimed at determining whether the 37-amino acid N-terminal tag altered the biophysical and pharmacological proper-ties of the Kv1.3 channel.
Voltage Dependence-Patch-clamp analyses of outside-out patches from dually-infected CV-1 cells at 12 or 24 h postinfection revealed large K ϩ -selective outward currents (Fig. 1). The outward currents appear to represent a single population of K ϩ channels that activate at depolarizing potentials with a V1 ⁄2 of Ϫ31 mV (Fig. 1A and Table I). This value is characteristic of Kv1.3 channels heterologously expressed in Xenopus oocytes (11) or mammalian cells (20), and of the native channels in mouse or human T-cells (30).
Deactivation and Inactivation-The time course of inactivation of recombinant Kv1.3 is similar to that of Kv1.3 expressed in mammalian cells (Table I). Like their native counterparts (3,11,20), the Kv1.3-pTH1 channels exhibit cumulative inactivation during 200 ms depolarizing pulses to ϩ40 mV from a holding potential of Ϫ80 mV, repeated once every second (Fig.  1, B and C). Channel deactivation provides another convenient property to distinguish between diverse K ϩ channel types. The kinetics of K ϩ channel closing can be determined by first opening the channels with a 15-ms conditioning pulse to ϩ40 mV and then forcing the channels to close by repolarizing to different potentials (Fig. 1D). The time constant tail of the resultant "tail" currents is similar to that of the native Kv1.3 channel (Table I).
Pharmacology-The pharmacological profile provides a further test for defining Kv1.3-pTH1 channels. As shown in Fig. 1, E and F, and in Table I, the Kv1.3-pTH1 channels, like their native Kv1.3 counterparts, are moderately sensitive to tetraethylammonium and highly sensitive to the peptide toxins ChTX, MgTX, and ShK (Table I). Our results indicate that the biophysical and pharmacological properties of the Kv1.3-pTH1 channels are indistinguishable from those of native Kv1.3 channels (11) and that the 37-amino acid N-terminal tag does not perceptibly alter channel function. By determining the peak current amplitudes and membrane capacitances of these patches and by comparing their membrane capacitances with those of whole cells, we estimate that between 1 ϫ 10 4 and 5 ϫ 10 5 Kv1.3 channels are expressed at the cell surface of CV-1 cells 24 h postinfection (data not shown).
Visualization of Kv1.3 Protein in CV-1 Cells by Immunofluorescence Microscopy-Kv1.3 protein is detectable in CV-1 cells at 4 h postinfection, and the intensity of staining increases significantly over the next 8 h (Fig. 2). Although the protein is widely distributed in the cell, staining is not visible at any time within the nucleus. A similar pattern is seen in RBL cells, although at 12 h the protein appears to be primarily localized at or near the cell surface (data not shown). The biophysical and immunofluorescence experiments clearly demonstrate that a substantial amount of Kv1.3 protein is produced by the heterologous VV-expression system. Single-step Purification of Membrane-associated Kv1.3 Protein-We solubilized Kv1.3 protein from the membrane fraction, but unlike the case for Shaker expressed in Sf9 cells, where a significant proportion of the protein was lost as insoluble aggregates in inclusion bodies (7-10), the majority of our protein was solubilized in nondenaturing detergent. The purified Kv1.3 protein was solubilized by CHAPS or cholate and stained poorly with silver, as has been reported for the Shaker protein (9, 10). Use of a higher concentration of SDS (7.5%) in the SDS-PAGE running buffer, however, enhanced the visualization of the protein in silver-stained gels. Ni 2ϩ -chelate chromatography greatly enriched a 64-kDa protein, visible on the silver-stained SDS-polyacrylamide gel (Fig. 3, lane 2), consistent with the expected size for Kv1.3. Immunoblotting with the anti-gene 10 Ab confirmed this protein to be Kv1.3 (Fig. 3, lane  3). Similar results were obtained for cholate-solubilized protein (data not shown). Table II shows results from one representative experiment. Total membrane protein was quantified by two independent methods which gave consistent results, namely the bicinchoninic assay and absorbance at 280 nm. Quantitative immunodot blots using anti-gene 10 antibody were used to estimate the relative yield of Kv1.3. From 115 mg of total solubilized membrane protein, we recovered 340 g of gene 10-reactive protein; this represents ϳ23% of the total immunologically reactive material in the solubilized membrane fraction and an enrichment factor of ϳ90-fold. These experiments were repeated several times (n Ͼ 10) with yields varying from 10 to 60 g/10 7 cells for CHAPS-or cholate-solubilized protein.

Kv1.3 Is Glycosylated and the Purified Protein Appears Relatively Homogeneous in This Respect-An N-glycosylation consensus site (NX[ST]
) is present in the S1-S2 extracellular loop of most Kv1 family proteins including Kv1.3 (2,9,10,(32)(33)(34); four additional N-glycosylation motifs are present in the N and C termini of Kv1.3, but since these are predicted to be located intracellularly they are unlikely to be utilized (2). To determine whether Kv1.3 is glycosylated in this expression system, we immunoprecipitated [ 35 S]methionine-labeled Kv1.3 from VVinfected CV-1 cells cultured in the presence or absence of tunicamycin, an inhibitor of N-linked glycosylation; the Kv1.3 analyzed here is derived from both intracellular and cell surface protein. Tunicamycin treatment resulted in a small (ϳ2-3 kDa) but reproducible shift in the migration of Kv1.3 in SDS-PAGE gels (Fig. 4, top); tunicamycin also inhibited incorporation of [ 14 C]glucosamine into Kv1.3 (data not shown). Treatment with N-glycosidase F of whole-cell solubilized protein from dually VV-infected cells, followed by Western blotting with the antigene 10 Ab to detect Kv1.3, revealed a similar reduction in size of the Kv1.3 protein (Fig. 4, bottom left). These data indicate that the purified Kv1.3 protein, both from intracellular and cell surface compartments, is relatively homogeneous and bears a single N-linked core sugar moiety which contributes 2-3 kDa to the mass of the protein. The faint smudge seen above the major 64-kDa band may represent Kv1.3 modified by other processes, for example phosphorylation (35).
Absence of the core sugar moiety resulting from tunicamycin treatment of CV-1 cells did not perceptibly alter the biophysical properties of Kv1.3 nor did it change expression levels (data not shown). As expected, the purified Kv1.3 protein also demonstrated the 2-3-kDa shift following treatment with N-glycosidase F (Fig. 4, bottom right), indicating that the majority of purified Kv1.3 is glycosylated as is its native counterpart in CV-1 cells.
The closely related Shaker channel is also glycosylated in its S1-S2 loop by at least two moieties that increase the mass of the protein by 3 and 6 kDa, respectively, and is functionally unaffected by deglycosylation (9). In contrast, voltage-gated sodium channels are extensively modified by carbohydrate, which accounts for 30% of their mass, and deglycosylation causes significant shifts in the voltage dependence of activation and enhances the frequency of reversible transitions to subconductance states (36).

Purified Kv1.3 Is a Tetramer with an Approximate Mass of 270 kDa and Dimensions of 65 ϫ 65 Å-Biophysical studies on
the Shaker channel have suggested that functional Kv proteins are tetrameric (37). Earlier EM studies on a single-particle preparation of Shaker protein revealed tetramers with dimensions of 80 ϫ 80 Å, although the mass of the complex was not determined (8). We have estimated the mass of our purified Kv1.3 protein by sedimentation and have utilized electron microscopy to visualize the protein in a near-crystalline array.
Purified Kv1.3 was centrifuged on a continuous 5-50% sucrose density gradient, and the resultant fractions were assayed for Kv1.3 by immunoblotting and quantitated by densitometry. The Kv1.3 protein, purified either in CHAPS (Fig. 5, top left) or in cholate (Fig. 5, top right), appeared almost exclusively as a single peak between the molecular weight standards apoferritin and alcohol dehydrogenase; no evidence was found for the presence of higher molecular weight aggregates. We estimated the molecular mass of Kv1.3 by the method of Martin and Ames (25), with the results shown in Fig. 5 (bottom). Since CHAPS and cholate detergent micelles are small (Ͻ10 kDa), their contribution to the mass of the complex is likely to be minimal. Our results suggest that the Kv1.3 complex is a 270-kDa oligomer, most likely a tetramer, made up of 64-kDa monomer subunits.
An additional demonstration of the tetrameric nature of Kv1.3 channels was obtained by visualizing the purified protein, reconstituted into liposomes, using negative staining electron microscopy. As shown in Fig. 6, the reconstituted protein forms small crystalline domains comprised of Kv1.   ized the biophysical properties of the purified protein (7-10). We therefore reconstituted purified Kv1.3 into planar lipid bilayers to ascertain the functional properties of the isolated channel (Fig. 7).
Single channel records measured at various potentials show the voltage dependence of the reconstituted channels (Fig. 7A). At negative voltages the probability of opening is significantly greater than at positive potentials, suggesting that the external side of the channel is located on the side of the bilayer to which vesicles were added (Fig. 7A). Thus, the K ϩ distribution in our lipid bilayer system is inverted, a high potassium concentration (250 mM) being present at the outer surface of the channel; negative voltages in the bilayer therefore correspond to depolarization in cells. The reconstituted channels begin to open at voltages negative to ϩ60 mV (see ϩ40 mV trace in Fig.  7A); in cells this would correspond to activation at voltages positive to Ϫ60 mV, similar to the Kv1.3 channel expressed in mammalian cells or Xenopus oocytes (11, 20; see Table I). Consistent with the channel's being highly K ϩ -selective, the reversal potential measured in asymmetric solutions (250 K o /25 K i ) was ϳ58 mV (Fig. 7B), and changed minimally when Na ϩ was added to the inside (250 K o /25 K i ϩ 225 Na i ; Fig. 7C).
As an independent test of function, we examined the reconstituted channel's sensitivity to the peptide toxins MgTX and ShK. Native Kv1.3 channels are blocked with high affinity by both these toxins (3,20,38), but toxin potency is reduced as the salt concentration increases, presumably due to electrostatic shielding or competition for a common binding site between K ϩ and the toxin. In order to maintain the stability of the bilayer, all the toxin binding experiments were performed at salt concentrations (100 K o /10 K i ) higher than those used in cell systems, making it necessary to use a higher external concentration of MgTX (300 nM) and ShK (100 nM) in our experiments. As shown in Fig. 7, D and E, the Kv1.3 currents in bilayers were completely blocked by MgTX and ShK. Taken together, our data indicate that the purified Kv1.3 protein reconstituted into lipid bilayers forms voltage-dependent K ϩ -selective channels that are potently blocked by these peptide toxins. These results extend earlier reports on the reconstitution of membrane fractions containing Shaker channels (33).
Since the probability of protein-carrying vesicle fusion with the bilayer is unknown, bilayer conductance cannot be used as a quantitative assay for the degree of purity of Kv1.3 protein.
However, reconstitution of membrane proteins nearly always involves the use of detergents that can destabilize the bilayer, a situation which dictates a narrow range of experimental conditions where the amount of protein added to the bilayer is adequate to observe channels but not enough to break the bilayer. If the purified material contained only a small fraction of functional Kv1.3 channels along with a large proportion of nonfunctional but membrane-associated proteins, adding sufficient material to observe currents would likely result in a broken bilayer or in a noisy record. Such noisy records or broken bilayers are not seen with any greater frequency after the addition of reconstituted Kv1.3 protein compared with control bilayers where nothing is added. Thus, we conclude that the reconstituted material is rich in functional Kv1.3 protein and is relatively free of contaminating, nonactive membraneassociated material. The fact that the protein appears to be uniformly glycosylated (Fig. 4) and assumes a largely uniform quaternary conformation (Figs. 5, 6) supports the idea that the protein may be functionally homogeneous as well. A more Electron micrograph of Kv1.3 protein reconstituted with dimyristoyl phosphatidylcholine at a low lipid-to-protein ratio. The image shows many small crystalline domains of Kv1.3 protein negatively stained with uranyl acetate (above). The well defined square-shaped objects seen at higher magnification (below), one of which is indicated by an arrow, appear to represent tetramers of Kv1.3. A centrally located stain-filled depression can be seen in many tetramers and is presumed to represent the location of the ion conduction pathway. Scale bars are shown for each image.
quantitative functional assay such as radiolabeled toxin binding was not possible, since the addition of CHAPS or cholate to VV-infected cell membranes containing Kv1.3, prior to any purification step, markedly reduced 125 I-MgTX binding, similar to the rapid decay in 125 I-ChTX binding reported with solubilized Shaker protein (7). Collectively, our data are consistent with a significant proportion of the purified protein being functional.
Conclusion-Using a modified VV-based system, we have successfully expressed 10 4 to 5 ϫ 10 5 functional Kv1.3 channels on the surface of each infected CV-1 cell, and the membraneassociated Kv1.3 protein is readily solubilized in nondenaturing detergents (either CHAPS or cholate). Reconstitution of purified protein into lipid bilayers produces Kv1.3-like currents; to our knowledge these experiments represent the first biophysical characterization of a purified Kv protein. By combining biochemical and electron microscopic approaches, we have demonstrated that the purified Kv1.3 protein complex has a mass of ϳ270 kDa and forms small crystalline domains in lipid membranes, consisting of well-defined tetramers with horizontal dimensions of 65 ϫ 65 Å; a central stain-filled density seen in these tetramers may represent the location of the channel pore. These results extend our earlier work using scorpion peptide toxins to map the pore of Kv1.3 (3,39), which revealed the existence of a 30-Å wide and 4 -6-Å deep external vestibule which narrows to ϳ9 Å at the external entrance to the ion conduction pathway. Future studies with two-dimensional crystals may provide a higher resolution structure of the Kv1.3 tetramer.