Kvβ Subunit Oxidoreductase Activity and Kv1 Potassium Channel Trafficking

Voltage-gated Kv1 potassium channels consist of pore-forming α subunits and cytoplasmic Kvβ subunits. The latter play diverse roles in modulating the gating, stability, and trafficking of Kv1 channels. The crystallographic structure of the Kvβ2 subunit revealed surprising structural homology with aldo-keto reductases, including a triosephosphate isomerase barrel structure, conservation of key catalytic residues, and a bound NADP+ cofactor (Gulbis, J. M., Mann, S., and MacKinnon, R. (1999) Cell 90, 943–952). Each Kv1-associated Kvβ subunit (Kvβ1.1, Kvβ1.2, Kvβ2, and Kvβ3) shares striking amino acid conservation in key catalytic and cofactor binding residues. Here, by a combination of structural modeling and biochemical and cell biological analyses of structure-based mutations, we investigate the potential role for putative Kvβ subunit enzymatic activity in the trafficking of Kv1 channels. We found that all Kvβ subunits promote cell surface expression of coexpressed Kv1.2 α subunits in transfected COS-1 cells. Kvβ1.1 and Kvβ2 point mutants lacking a key catalytic tyrosine residue found in the active site of all aldo-keto reductases have wild-type trafficking characteristics. However, mutations in residues within the NADP+ binding pocket eliminated effects on Kv1.2 trafficking. In cultured hippocampal neurons, Kvβ subunit coexpression led to axonal targeting of Kv1.2, recapitulating the Kv1.2 localization observed in many brain neurons. Similar to the trafficking results in COS-1 cells, mutations within the cofactor binding pocket reduced axonal targeting of Kv1.2, whereas those in the catalytic tyrosine did not. Together, these data suggest that NADP+ binding and/or the integrity of the binding pocket structure, but not catalytic activity, of Kvβ subunits is required for intracellular trafficking of Kv1 channel complexes in mammalian cells and for axonal targeting in neurons.

Voltage-dependent potassium channels of the Shaker or Kv1 family play a fundamental role in the mammalian nervous system by determining resting membrane potential, frequency of action potential firing, and neurotransmitter release (2). Mammalian Kv1 channels are composed of four voltage-sensing and pore-forming transmembrane Kv1 ␣ subunits, and up to four cytoplasmic Kv␤ subunits (3). A tremendous diversity in Kv channel structure arises from the combinatorial assembly of the products of 18 ␣ subunit and 4 ␤ subunit genes present in the mammalian genome (4). Moreover, different Kv1 ␣ and Kv␤ subunit combinations yield functionally distinct potassium channels (5). Different Kv␤ subunits, namely Kv␤1.1, Kv␤1.2, Kv␤2, and Kv␤3, have distinct effects on channel function, most obviously on fast inactivation because of the "ball" inactivation peptide present on the N terminus of some but not all Kv␤ subunits (6). The distinct N-terminal domains also mediate subtypespecific interaction with the cytoskeleton (7), which can further impact Kv␤ subunit effects on inactivation (8).
Outside of these distinct N-terminal domains, the primary sequence of Kv␤ subunits is highly conserved, such that all Kv␤ subunits exhibit Ͼ85% amino acid identity over this Ϸ330-amino acid "core" region. Analysis of the deduced amino acid sequence of this core domain led to the proposal (9) that Kv␤ subunits may be members of the aldo-keto reductase (AKR) 1 enzyme family, which reduce aldehyde or ketone functional groups to primary or secondary alcohols. The crystal structure of the Kv␤2 core domain extended this analysis and revealed remarkable structural similarity to AKR family members (1). The core domain of Kv␤2 consists of a 4-fold symmetrical triosephosphate isomerase barrel structure with bound NADP ϩ cofactor. The essential features of catalysis by AKRs involve hydride transfer from the nicotinamide ring to a carbonyl carbon of the substrate, followed by proton transfer from a donor group on the enzyme to the carbonyl oxygen of the substrate. The Kv␤2 structure (1) predicts that Tyr-90 acts as the proton donor of Kv␤2 ( Fig. 1) in analogy to the Tyr-48 residue of human aldose reductase (10). Residues Trp-57, Arg-189, Trp-243, Ser-244, and Arg-264 are important for creating the NADP ϩ binding pocket (Fig. 1). Despite the striking structural homology between Kv␤2 and bona fide AKRs, it has not been demonstrated experimentally that Kv␤ subunits in fact have enzymatic activity and if so, the identity of their physiological substrates. One subsequent study has revealed that catalytic and cofactor binding site mutants are deficient in their ability to confer N-type inactivation to Kv1.5, suggesting that Kv␤ AKR activity and Kv1 channel inactivation may be coupled (11).
The vast majority of Kv1 channel complexes in mammalian brain have associated Kv␤2 subunits (12)(13)(14), which do not confer N-type inactivation on associated Kv1 channels (6). The interaction of Kv1 ␣ and Kv␤ subunit polypeptides is an early event in Kv1 channel biosynthesis, occurring in the endoplasmic reticulum (ER) (15,16). This interaction is quite stable, and subsequent subunit exchange is not observed (15,17). The * This work was supported by National Institutes of Health Grant NS34383, by Wyeth-Ayerst Research, and by the Center for Biotechnology at Stony Brook, funded by the New York State Science and Technology Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: interaction with Kv␤2 subunits results in increased stability of Kv1.2 ␣ subunits and effects on cotranslational addition of N-linked oligosaccharides to Kv1.2 which are consistent with more efficient ␣ subunit folding in the ER (15). Increased surface expression of Kv1 ␣ subunits was also observed upon Kv␤2 coexpression, as revealed by staining intact cells with external antibodies (15,18), 125 I-labeled dendrotoxin binding (15), and clustering of cell surface channels by PSD-95 (19). Similar Kv␤2-mediated effects on surface expression were observed for Kv1.1-Kv1.2 heteromeric channels (20). Increases in the amplitude of potassium currents in Xenopus oocytes expressing certain Kv1 ␣ subunits have also been observed upon Kv␤2 coexpression (21). These studies together imply that in addition to the role of Kv␤ subunits as components of plasma membrane Kv channel complexes, Kv␤ subunits may play a fundamental biosynthetic role in regulating intracellular trafficking and surface expression of Kv channels.
In this study, we first addressed whether all members of the Kv␤ subunit gene family, each of which contains the highly conserved core region with key catalytic site and cofactor binding site residues, could affect Kv1.2 channel intracellular trafficking and surface expression as was observed previously for Kv␤2 (15). Despite dramatic differences in their effects on channel gating, each of the Kv␤ subunits displayed robust trafficking effects. A recent study has shown that a double mutation (D85A/Y90F) in the Kv␤2 catalytic domain reduces the ability of Kv␤2 to promote plasma membrane trafficking of Kv1.4 in Xenopus oocytes (22). Here we address further the role for AKR activity in Kv1 channel trafficking by mutating only the side chain hydroxyl group of this critical catalytic tyrosine residue in Kv␤1.1 (Y124F) and Kv␤2 (Y90F), and using a variety of biochemical and cell biological assays to analyze the effects of this mutation on Kv1.2 trafficking in mammalian cells. We also investigate the role of key cofactor binding site residues in Kv␤-mediated trafficking of Kv1.2 through structure-based mutagenesis. Finally, we show that Kv␤ subunits that enhance cell surface expression of Kv1.2 in COS-1 cells can also induce the targeting of Kv1.2 to axons in cultured hippocampal neurons, recapitulating the subcellular localization of Kv1.2 observed in mammalian central and peripheral neurons.

EXPERIMENTAL PROCEDURES
Materials-All materials not specifically identified were purchased from Sigma or Roche Molecular Biochemicals.
Generation of Kv␤ Point Mutants-Point mutants were generated using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) as per the manufacturer's instructions.
Analyses of Transfected COS-1 Cells-Cells were transfected with mammalian cDNA expression vectors for various Kv channel ␣ and ␤ subunit polypeptides (26) by the calcium phosphate precipitation method (27). Cells expressing various amounts of Kv1 ␣ and Kv␤ subunits were stained 48 h post-transfection using either a standard immunofluorescence staining protocol for permeabilized cells (27) or a surface immunofluorescence staining protocol for intact cells (15,20). Cells were viewed under indirect immunofluorescence on a Zeiss Ax-ioskop2 microscope. Surface versus total immunofluorescence staining was scored under narrow wavelength fluorescein and Texas Red filter sets. Surface expression index (SEI) values were determined as the percentage of Kv1.2-transfected cells with Kv1.2 surface staining and represent the mean Ϯ S.D. determined from 100 transfected cells from each of three dishes. Fluorescent images of cells were captured into Adobe Photoshop from either a SPOT (Diagnostic Instruments, Sterling Heights, MI) or a Zeiss Axiocam (Oberkochen, Germany) cooled CCD 24-bit color digital camera mounted on a Zeiss Axioskop 2 microscope with a 40ϫ, 1.3 numerical aperture plan neofluar objective, or a 63 ϫ 1.25 numerical aperture plan neofluar objective, using the software supplied with the cameras.
Confocal images were generated on a Zeiss LSM510 laser scanning confocal microscope system with an Axiovert microscope using a 63 ϫ 1.25 numerical aperture plan neofluar objective. Alexa 594 (red) immunofluorescence signals were obtained using a 560 nm long pass filter after excitation with a helium neon laser at 543 nm. Alexa 488 (green) immunofluorescence signals were obtained using a 505-530 band pass filter after excitation with an argon laser at 488 nm. Red and green signals were generated and collected individually on a frame-by-frame basis. Similar pinhole sizes and amplifier settings were used to obtain all images, and no further manipulation of image files was performed. Each individual image represents a three-dimensional projection of the entire cell derived from a z-series taken as an average of eight sweeps at 1024 ϫ 1024 resolution and viewed en face relative to the apical surface. Projections were exported as Photoshop files for presentation.
SDS-PAGE and immunoblotting analyses of cell lysates prepared from transfected cells were performed as described (15,20). For proteinase K digestion, transfected cells were washed three times with ice-cold phosphate-buffered saline. Each 35-mm dish was incubated with 10 mM HEPES, 150 mM NaCl, and 2 mM CaCl 2 (pH 7.4) with or without 200 g/ml proteinase K (28) at 37°C for 30 min. The cells were then harvested and centrifuged at 4°C at 3,500 rpm in a refrigerated microcentrifuge, and proteinase K digestion was quenched by adding ice-cold phosphate-buffered saline containing 6 mM phenylmethylsulfonyl fluoride and 25 mM EDTA. This was followed by three washes in ice-cold phosphate-buffered saline. Cell lysates were prepared and analyzed by immunoblot as described above.
Primary Hippocampal Cultures-Hippocampal cultures were prepared as described previously (29,30) with modifications (31). Cover- slips were placed in wells face up but did not contact the glia because of the presence of paraffin wax pedestals. Most (Ͼ85%) of the neurons in these cultures developed the characteristic morphology of hippocampal pyramidal cells. Cultured hippocampal neurons at 7 days in vitro were transfected by the LipofectAMINE Plus method (Invitrogen) as described previously (31). In brief, 1 g of DNA was diluted into 100 l of serum-free transfection medium to which 10 l of PLUS reagent was added, the solution was mixed and incubated at room temperature for 15 min. Then 5 l of LipofectAMINE reagent was diluted into serumfree dilution medium in a second tube and mixed. Precomplexed DNA and diluted LipofectAMINE reagent were combined, mixed, and incubated for 15 min at room temperature. The DNA-PLUS-LipofectAMINE reagent complexes were added to face up coverslips in individual wells of six-well tissue culture plates. Complexes were mixed into the medium gently and incubated at 37°C with 5% CO 2 for 48 h. Cells expressing various amounts of Kv1.2 ␣ and Kv␤ subunits were stained 48 h post-transfection using a standard immunofluorescence protocol (20,27,31). Cells were viewed under indirect immunofluorescence on a Zeiss Axioskop2 microscope, as described above. Axonal versus total Kv1.2 staining was scored under narrow wavelength fluorescein and Texas Red filter sets. Axonal Expression Index values were determined as the percentage of transfected cells with Kv1.2 staining of fine caliber MAP2(Ϫ) processes (axons) and represent the mean Ϯ S.D. determined from 50 transfected cells from each of three dishes.

All Kv␤ Subunits Promote Surface Expression of Kv1.2 ␣
Subunits-All Kv␤ subunits have a highly conserved Ϸ330amino acid core domain that follows divergent N termini of 30 -70 amino acids. Previous studies have shown that coexpression with either Kv␤1.1 or Kv␤2 increased 125 I-dendrotoxin binding in cells expressing Kv1.2 (15) and that the autonomous highly conserved core domain is sufficient to mediate increases in current amplitude of coexpressed Kv1.4 ␣ subunits (21). We addressed whether each wild-type Kv␤ subunit harboring this core domain would display the effects on Kv1.2 intracellular trafficking and surface expression observed previously for Kv␤2 (15). Rat Kv␤1.1 and Kv␤2 (26,32), human Kv␤1.2 (33), and rat Kv␤3 (34) were coexpressed with the rat Kv1.2 ␣ subunit in COS-1 cells. We performed double immunofluorescence staining using antibodies specific for Kv1.2 extracellular (prior to detergent permeabilization) and intracellular (after detergent permeabilization) domains to determine the relative amount of Kv1.2 surface expression in the transfected cells. In the absence of any Kv␤ subunit coexpression, the majority of Kv1.2-expressing cells exhibited robust perinuclear Kv1.2 staining, which we have previously shown corresponds to ERlocalized Kv1.2 (20), and no detectable surface staining ( Fig The deduced amino acid sequence of the Kv1.2 polypeptide contains a single N-linked glycosylation site located in the extracellular loop between transmembrane segments S1 and S2 (4). Differences in processing of N-linked oligosaccharide chains at this site can be used as an independent biochemical marker of intracellular trafficking via shifts in apparent molecular mass on SDS-PAGE gels (20,35). Lower (Ϸ60 kDa) molecular mass forms of Kv1.2 carry simple high mannose chains and correspond to ER pools, whereas higher (Ϸ86 kDa) molecular mass forms, similar to Kv1.2 in rat brain membranes, carry sialylated complex chains and correspond to Golgi and plasma membrane pools (20). We found the bulk of Kv1.2 present in COS-1 cells expressing Kv1.2 alone in the lower molecular mass ER form (Fig. 2C) (20). However, coexpression with Kv␤2 yields a dramatic increase in amount of Kv1.2 in the higher molecular mass post-ER form, similar to the Kv1.2 found in rat brain (Fig. 2C). Coexpression with other Kv␤ subunits yielded similar increases in the amount of Kv1.2 that had undergone post-ER processing of the N-linked oligosaccharide chain (data not shown). Together, these cell biological and biochemical analyses reveal that coexpression with Kv␤ subunits increases the cell surface expression of Kv1.2 ␣ subunits with a concomitant decrease in retained ER pools.
Effect of Kv␤2 Catalytic Site Mutants on Kv1.2 Surface Expression-Previous studies on AKR family members have revealed a critical role for the hydroxyl group of a specific catalytic site tyrosine in proton transfer to the substrate carbonyl oxygen (10). The crystal structure of Kv␤2 (1) defines Tyr-90 as this residue in Kv␤2 (Fig. 1). Mutation of the analogous tyrosine (Tyr-48) in human aldose reductase to phenylalanine (Y48F), which preserves both the spatial fit and hydrophobicity of the tyrosine residue yet eliminates the possibility of any general acid catalysis by removing the tyrosine hydroxyl group, led to a completely inactive enzyme (10). To determine the role of the potential Kv␤ subunit enzymatic activity in Kv1 channel trafficking experimentally, we made the analogous tyrosine to phenylalanine mutation (Y90F) in Kv␤2. A less subtle Kv␤2 Y90A mutation was also made.
When coexpressed with Kv1.2 in COS-1 cells at a fixed ␣/␤ subunit cDNA ratio of 1:4, the Kv␤2 Y90F mutant was indistinguishable from wild-type Kv␤2 in its effects on promoting increased surface expression of Kv1.2 (Fig. 3A). In contrast, the Kv␤2 Y90A mutant failed to induce increased surface expression of Kv1.2 and was not significantly different from no Kv␤ subunit addition (Fig. 3A). In these experiments the steadystate expression level of wild-type Kv␤2 and the two different mutant proteins was similar (data not shown). Moreover, dose-response curves show that the respective phenotypes of the wild-type and mutant Kv␤2 isoforms shown in Fig. 3A are maintained over an 8-fold range (1:1-1:8) of ␣/␤ subunit cDNA ratios (data not shown). As such, the lack of activity of Kv␤2 Y90A is not simply the result of differences in expression level.
To confirm and extend these cell biological assays, quantitative biochemical assays of Kv1.2 cell surface expression were performed. Cells expressing Kv1.2 with and without coexpressed wild-type and mutant Kv␤2 subunits were treated with proteinase K, a relatively nonspecific protease that when externally applied cleaves the ectodomains of cell surface proteins (28). Cell extracts prepared from proteinase K-treated and control cultures were then analyzed by SDS-PAGE and immunoblotting. We have established previously the utility of this assay in analyses of trafficking and surface expression of Kv1 ␣ subunits (20,36). As shown in Fig. 3B and 3B). In each case treatment of the cultures with externally applied proteinase K eliminated the bulk of the 86-kDa form of Kv1.2 (Fig. 3B), verifying the cell surface localization of this polypeptide pool. Digestion with sialidase confirmed that the higher apparent molecular mass pool in each sample carried sialic acid (data not shown) because of the processing of the N-linked oligosaccharide chains on this Golgi/plasma membrane pool of Kv1.2. Together these data show that the Kv␤2 Y90F mutation, which, based on previous studies, should eliminate any AKR activity of Kv␤2, is phenotypically identical to wild-type Kv␤2 in its effects on Kv1.2 trafficking, whereas Kv␤2 Y90A exhibits little or no trafficking activity.
When expressed in COS-1 cells, both wild-type Kv␤2 (7) and the Kv␤2 Y90F mutant exhibit a diffuse cytoplasmic localization (Fig. 4A). However, we found that the Kv␤2 Y90A mutant is present as perinuclear clusters suggestive of a misfolded phenotype (Fig. 4B). Moreover, the altered localization of Kv␤2 Y90A is conferred to coexpressed Kv1.2 (Fig. 4B). These results suggest that the Kv␤2 Y90A mutant retains its ability to interact with Kv1.2 but that this interaction does not allow for enhanced surface expression of Kv1.2. To determine directly whether the lack of ability of Kv␤2 Y90A to promote Kv1.2 surface expression was not simply the result of a lack of association, we performed coimmunoprecipitation experiments. As shown in Fig. 4C, each of the wild-type and mutant Kv␤2 subunits is found associated with Kv1.2, although the interaction of Kv1.2 with the Kv␤2 Y90A mutant is slightly less robust. Thus, the lack of ability of Kv␤2 Y90A to promote the surface expression of Kv1.2 is not simply the result of a lack of interaction.

Effect of Cofactor Binding Site Mutations on Kv1.2 Surface Expression-
The other domain important in the potential AKR activity of Kv␤ subunits is the binding site for the NADP ϩ / NADPH cofactor (Fig. 1). NADPH cofactor would serve as the potential electron donor for substrate reduction. NADP ϩ was present in the crystallographic structure of bacterially expressed Kv␤2 (1), and residues lining the cofactor binding pocket are 100% conserved between Kv␤1.1 and Kv␤2 (32). We addressed the role of cofactor binding in Kv␤ subunit function by analyzing Kv␤1.1 mutants carrying structure-based point mutations in the cofactor binding region (Fig. 1) for their ability to promote surface expression of Kv1.2. Mutations were made in 10 different residues lining the cofactor binding site, and in each case more than one substitution was made for each residue. When expressed in COS-1 cells, the subcellular distribution of each of these mutants was similar to wild-type Kv␤1.1 and consisted of staining associated with the actin cytoskeleton (7) and additional perinuclear staining (Fig. 5A). Thus, unlike Kv␤2 Y90A, none of the cofactor binding site mutations in Kv␤1.1 led to a grossly altered subcellular distribution.
Surface staining of transfected COS-1 cells expressing Kv1.2 and each of the Kv␤1.1 cofactor binding site mutants revealed that each of these mutants had lost its ability to promote Kv1.2 surface expression (Fig. 5B). Coexpression with any of the cofactor binding site mutants was not significantly different from no Kv␤ subunit addition (Fig. 5B). The nature of the mutation (e.g. charged to uncharged, polar to nonpolar, etc.) did not significantly affect the extent of activity loss (e.g. compare K288A with K288D). These results suggest that cofactor binding and/or the structure of the cofactor binding domain is critical to the ability of Kv␤1.1 to promote the surface expression of Kv1.2.
We next used structure-based site-directed mutagenesis to investigate the role of three other functional domains of Kv␤1.1 in Kv1.2 trafficking. Mutations were made in key residues for binding to the T1 domain of Kv1 ␣ subunits (37), in the substrate binding pocket (1), and in the catalytic domain (1). Mutation of Kv␤1.1 Arg-237 (Arg-203 in Kv␤2), present at the ␣/␤ subunit interface (37), to either Ala or Phe eliminates the ability of Kv␤1.1 to promote the surface expression of Kv1.2 (Fig. 5B). Similar loss of activity was observed upon mutation of Kv␤1.1 Trp-155 (Trp-121 in Kv␤2; Fig. 1), which is predicted to line the substrate binding pocket (1). Mutations were also made in the catalytic Tyr residue (Kv␤1.1 Tyr-124) structurally analogous to Kv␤2 Tyr-90. Similar to the results obtained with Kv␤2 Tyr-90, the Kv␤1.1 Y124F mutant had wild-type characteristics. The Kv␤1.1 Y124H mutation, however, eliminated the ability of Kv␤1.1 to promote surface expression of Kv1.2 in COS-1 cells. Together these data show that mutations at diverse sites (cofactor, ␣ subunit, and substrate binding) in the Kv␤1.1 core domain result in loss of trafficking activity. However, mutation of a critical hydroxyl group from a key catalytic tyrosine residue does not affect the trafficking activity of either Kv␤1.1 or Kv␤2, although more drastic alterations of the side chain do reduce activity.
Effects of Kv␤ Subunits on Kv1.2 Trafficking in Hippocampal Neurons-Kv1.2 ␣ subunits are found associated with Kv␤2 subunits on axons and terminal fields in the mammalian central and peripheral nervous systems (12,14,38). We next used transfection of recombinant potassium channel subunits into low density cultures of embryonic rat hippocampal neurons to address the role of Kv␤2 subunits in axonal targeting of Kv1.2. These cultures do not express detectable Kv1.2 or Kv␤2 staining at the age of culture used for these experiments (20). Hippocampal neurons transfected with Kv1.2 alone express robust Kv1.2 staining associated with the soma and very proximal portions of the dendrites (Fig. 6A), suggesting that as in COS-1 cells, the majority of Kv1.2 is found in the ER. A similar staining pattern was obtained with antibodies specific for BiP, a resident ER protein (not shown). Hippocampal neurons transfected with wild-type Kv␤2 alone express robust Kv␤2 staining in the soma, dendrites, and in the MAP2(Ϫ) fine caliber axon (Fig. 6B). Given the lack of axonal staining of introduced Kv1.2, we next addressed whether Kv␤2 subunit coexpression could induce the appropriate trafficking of Kv1.2 to axons. Hippocampal neurons cotransfected with Kv1.2 and Kv␤2 exhibited robust Kv1.2 staining on axons which was not observed in cells expressing Kv1.2 alone (Fig. 6C). Kv1.2 staining was also observed on the soma and proximal dendrites (Fig. 6C), presumably representing the biosynthetic Kv1.2 pool. Double staining for dendritic MAP2 verified that the long, fine caliber processes with Kv1.2 staining in neurons coexpressing Kv␤2 were in fact MAP2(Ϫ) axons (not shown). Quantitative analyses of transfected neurons revealed a dramatic difference in the extent Kv1.2 axonal staining between neurons expressing Kv1.2 alone and those expressing both Kv1.2 and Kv␤2 (Fig. 6E). These data indicate that not only is coexpression with Kv␤2 necessary for robust Kv1.2 surface expression in COS-1 cells, but it is also required for the appropriate axonal targeting of Kv1.2 in hippocampal neurons.
We next addressed the effect of Kv␤2 catalytic site mutations on the axonal targeting of Kv1.2. Neurons cotransfected with Kv1.2 and Kv␤2 Y90F exhibited Kv1.2 staining on axons similar to that observed for wild-type Kv␤2 (data not shown). Moreover, Kv␤2 Y90F was almost as potent at inducing axonal localization of Kv1.2 as was wild-type Kv␤2 (Fig. 6E). Surprisingly, Kv1.2 was also localized to axons in Ϸ 30% of cells coexpressing Kv␤2 Y90A (Fig. 6E), suggesting that this particular mutant retains more trafficking activity in neurons than it does in COS-1 cells. However, a subset of neurons had Kv␤2 Y90A staining in large clusters in the soma (Fig. 6D), similar to what was observed in COS-1 cells expressing Kv␤2 Y90A (see Fig. 4A), and these neurons lacked axonal Kv1.2 staining (Fig. 6D).
We also analyzed the effects of wild-type Kv␤1.1 on axonal targeting of Kv1.2. Like Kv␤2, wild-type Kv␤1.1 coexpression led to robust targeting of Kv1.2 to axons (Fig. 6E). Axonal targeting was observed in far fewer cells when Kv1.2 was coexpressed with Kv␤1.1 W277V (Fig. 6E), a cofactor binding site mutant that did not promote Kv1.2 surface expression in COS-1 cells (Fig. 5B). Together these data suggest that an intact NADP ϩ binding site, but not catalytic activity per se, is required for efficient trafficking of Kv1.2 in COS-1 cells and in neurons.

DISCUSSION
Voltage-gated sodium, calcium, and potassium channels have a distinct repertoire of auxiliary subunits that can influence the gating, turnover rates, and intracellular trafficking of associated pore-forming and voltage-sensing transmembrane ␣ subunits (39). In particular, the cytoplasmic ␤ subunits of voltage-gated calcium and potassium channels have been shown to exert dramatic effects on the surface expression levels of associated ␣ subunits (40). Previous structure-function analyses of Kv␤ subunits have focused on domains of the polypeptide responsible for effects on inactivation (6,32,37,41) and on interaction with ␣ subunits (26,42,43) and cytoskeleton (7). However, unlike other auxiliary ion channel subunits, the crystallographic structure of the mammalian Kv␤2 subunit reveals a high degree of structural homology with AKRs (1), as had been predicted by earlier analyses of Kv␤ subunit sequences (9,44). Here we addressed the role of the putative enzymatic activity of Kv␤ subunits in intracellular trafficking of Kv1 ␣ subunits. We show that mutations in the cofactor binding site disrupt the ability of the Kv␤1.1 subunit to promote surface expression and axonal targeting of Kv1.2. However, discrete mutation of a key catalytic residue in Kv␤1.1 or Kv␤2 which would eliminate enzymatic function does not alter the Kv␤ trafficking phenotype. The overall conclusion drawn from these studies is that should Kv␤ subunits possess enzymatic activity, this activity is not necessary for their effects on intracellular trafficking, surface expression, and axonal targeting of Kv1.2 ␣ subunits.
Our previous studies showed that both Kv␤1.1 and Kv␤2 coexpression increased surface dendrotoxin binding activity of Kv1.2-expressing COS-1 cells (15). Here we show that all identified Kv1-associated Kv␤ subunits can promote increased surface expression of Kv1.2. The precise mechanism by which this is achieved remains to be elucidated. However, biochemical analyses presented here show that the relative proportions of Kv1.2 carrying the ER (high mannose) and post-ER (complex) forms of N-linked oligosaccharide chains are altered dramatically upon Kv␤ subunit coexpression. This indicates that the effects of Kv␤ subunits in increasing Kv1.2 surface expression are most probably the result of changes in the subcellular distribution of Kv1.2, as opposed to uniformly increasing expression levels with a resultant increase in cell surface levels. Similar results have been observed for Kv␤2 effects on Kv1.4 expression in Xenopus oocytes (22). Whether the enhanced surface expression of Kv1.2 results from more efficient export of Kv1.2 from the ER or enhanced stability of cell surface Kv1.2 is not known. A recent study described an autonomously acting ER export signal (FCYENE) that, when appended to the cytoplasmic C terminus of Kv1.2, increased Kv1.2 surface expression (18). Interestingly, this appended ER export signal oc- cluded the ability of Kv␤2 to promote Kv1.2 surface expression further (18), suggesting that the mechanism that underlies the Kv␤2-induced increases in Kv1.2 surface expression is the enhanced ER export of Kv1.2.
Previous studies of the enzyme kinetics and crystal structure of human aldose reductase (10) revealed a complete loss of enzymatic activity upon mutation of the tyrosine residue (Tyr-48) structurally homologous to Kv␤2 Tyr-90 and Kv␤1.1 Tyr-124. These and other studies of AKRs (45,46) have shown that the tyrosine hydroxyl group, whose pK a is perturbed through association with a neighboring lysine and aspartate, acts as the proton donor during the protonation of the carbonyl oxygen group of the substrate. Thus, eliminating the critical side chain hydroxyl group in the Kv␤2 Y90F and Kv␤1.1 Y124F mutants, although retaining the overall hydrophobicity and size of the residue, should eliminate catalysis in Kv␤s as it does in other AKRs. The complete lack of effects of the Kv␤2 Y90F and Kv␤1.1 Y124F mutations on any measurable aspect of Kv1.2 trafficking in COS-1 cells, and on axonal targeting in neurons, argues strongly against a role for AKR catalytic activity in these Kv␤ subunit-mediated events. It should also be noted that the Kv␤1.1 Y124F mutation also had no effect on the ability of Kv␤1.1 to confer N-type inactivation to Kv1.5 (11). However, a catalytic site double mutation in Kv␤2 (D85A/ Y90F) led to an observed loss of the Kv␤2-mediated enhancement of Kv1.4 plasma membrane trafficking in Xenopus oocytes (22). We have shown that Kv1.2 and Kv1.4 have inherently distinct trafficking characteristics in mammalian cells (20) because of differences in the sequence of a trafficking determinant located near the external mouth of the channel pore (36). Whether the differences between the findings of Peri et al. (22) and those presented here are the result of these inherent differences between Kv1.2 and Kv1.4 trafficking, differences between the oocyte and mammalian expression backgrounds, or unintentional effects of the Y90F/D85A double mutation on Kv␤2 structure, with a subsequent loss of Kv␤2 trafficking effects, is not known.
Given our results on the apparent lack of a requirement of catalytic activity for Kv␤-mediated trafficking effects, our observations on the dramatic and consistent effects of mutations in the NADPH cofactor binding site are somewhat surprising. The crystal structure of bacterially expressed Kv␤2 contained tightly bound NADP ϩ that presumably bound in vivo and remained tightly bound throughout the extensive purification and crystallization procedures (1). Having NADP ϩ /NADPH bound may be critical to the initial folding and/or stability of the Kv␤ subunit structure such that subunits without cofactor misfold and are degraded. A recent study showed that bound cofactor was critical to the proper folding of 2,5-diketo-D-gluconic acid reductase A, another member of the AKR family, with regions of the enzyme undergoing coordinated conformational changes of up to 8 Å when synthesized in the absence of cofactor (47). Fluorescence measurements have revealed that bacterially expressed Kv␤2 exhibits high affinity binding to both NADPH and NADP ϩ (48). A mutation in Kv␤2 analogous to our Kv␤1.1 cofactor binding mutation R298D (Kv␤2 R264E) resulted in a complete loss of NADPH binding (48). Interestingly, mutations at Kv␤2 residues analogous to Kv␤1.1 Arg-223, Trp-277, and Cys-282 did not affect NADPH binding (48), although we observed a loss of Kv␤1.1-induced trafficking effects. In only one case was the same mutation (W277A) analyzed, thus some of these differences may result from the differences in specific side chain substitution used in the two different studies. However, it is also possible that mutations in this highly conserved core domain induce structural alterations independent of cofactor binding which lead to loss of Kv␤1.1-mediated trafficking. Future analyses of NADP ϩ /NADPH binding in our mutants which display altered trafficking of Kv␤ subunit-associated complexes may help resolve these intriguing discrepancies.
The lack of effects of the discrete catalytic site mutations on the trafficking function of Kv␤ subunits raises questions as to why each of the detailed structural features of AKR family members is present in the Kv␤2 structure, and why all of the important domains are so highly conserved in all Kv␤ subunits. Certainly catalysis could be coupled to another aspect of Kv␤ subunit function, such as modulation of N-type inactivation (11), although this does not explain the conservation of AKR structure in Kv␤2. Another possibility is that Kv␤ subunits are "moonlighting" as bona fide AKRs in contexts outside their role as potassium channel subunits. This would be analogous to the enzymatic activity of gephyrin, a protein that associates with and clusters glycine receptor ion channels in neurons, but which in non-neuronal cells functions as an enzyme in molybdenum cofactor biosynthesis (49). A number of studies have suggested that to conserve genome size the same protein may be used for completely unrelated functions in different cells, depending on cell-specific protein-protein interactions affecting subcellular localization or quaternary structure (50). Because Kv␤ subunits have a more widespread tissue expression than do Kv1 ␣ subunits (NCBI Unigene), the catalytic activity of Kv␤2 may only be relevant in cells lacking Kv1 ␣ subunits and where Kv␤s function autonomously. Any physiological role of the putative catalytic activity of Kv␤ subunits remains to be elucidated.
Our findings that Kv␤ subunits mediate the trafficking of Kv1.2 to axons of cultured hippocampal neurons suggest a new role for cytoplasmic Kv␤ subunits in Kv channel function. Immunoelectron microscopy studies (51) and experiments utilizing circumscribed excitotoxic lesions (14) have shown that Kv1.2-containing Kv1 channels are associated predominantly with axons and axon terminals in the mammalian nervous system. Moreover, these axonal Kv1 channels are associated extensively with Kv␤ subunits, especially Kv␤1.1 and Kv␤2 (12,14). Because Kv␤2 coexpression leads to axonal targeting of Kv1.2 not observed in cells expressing Kv1.2 alone, it is interesting to speculate that axonal targeting signals may exist on Kv␤2 which are critical to Kv1 channel localization in neurons. Although determinants of polarized and clustered localization of dendritic Kv2.1 potassium channels in both polarized epithelial Madin-Darby canine kidney cells (52) and cultured hippocampal neurons (31) have been characterized, similar studies of signals involved in the localization of axonal potassium channels have not been performed. A previous study of the localization of recombinant Kv1.4 ␣ subunits expressed in Madin-Darby canine kidney cells revealed a basolateral localization (53). Based on the analogous membrane hypothesis (54), axonal Kv1.4 should assume an apical localization in Madin-Darby canine kidney cells, thus the predicted localization was not obtained. Another study using biolistic gene transfer to express recombinant Kv1.4 in high density postnatal hippocampal cultures also found an aberrant somatodendritic localization for Kv1.4 (55). It should be noted that both of these studies were performed in the absence of Kv␤ subunit coexpression. Future studies in cells coexpressing Kv1 ␣ and Kv␤ subunits may allow for a better understanding of the respective role of these component subunits in determining the subcellular localization of axonal potassium channels in neurons.