Genetic Analysis of the Mammalian K+ Channel β Subunit Kvβ2 (Kcnab2)*

Kvβ2 binds to K+ channel α subunits from at least two different families (Kv1 and Kv4) and is a member of the aldo-ketoreductase (AKR) superfamily. Proposed functions for this protein in vivo include a chaperone-like role in Kv1 α subunit biogenesis and catalytic activity as an AKR oxidoreductase. To investigate the in vivo function of Kvβ2, Kvβ2-null and point mutant (Y90F) mice were generated through gene targeting in embryonic stem cells. In Kvβ2-null mice, Kv1.1 and Kv1.2 localize normally in cerebellar basket cell terminals and the juxtaparanodal region of myelinated nerves. Moreover, normal glycosylation patterns are observed for Kv1.1 and Kv1.2 in whole brain lysates. Thus, loss of the chaperone-like activity does not appear to account for the phenotype of Kvβ2-null mice, which include reduced life spans, occasional seizures, and cold swim-induced tremors similar to that observed in Kv1.1-null mice. Mice expressing Kvβ2, mutated at a site (Y90F) that abolishes AKR-like catalytic activity in other family members, have no overt phenotype. We conclude that Kvβ2 contributes to regulation of excitability in vivo, although not directly through either chaperone-like or typical AKR catalytic activity. Rather, Kvβ2 relies upon as yet unidentified mechanisms in the regulation of K+ channel and/or oxidoreductive functions.

Kv␤2 was isolated biochemically through its association with the Kv1 family of K ϩ channels in a 4:4 stoichiometry (1) and later found to coassemble with Kv4 channels (2,3). It is expressed abundantly in the nervous system and also in T-lymphocytes, where it appears to play a role in K ϩ channel activation in at least some mammalian species (4). Subsequent cloning of the Kv␤2 homologs, Kv␤1and Kv␤3, revealed a conserved core region and variable N-terminal domains that may contain a ball-and-chain motif conferring rapid inactivation to even noninactivating Kv1 ␣ subunits (5)(6)(7). The observed alterations in K ϩ channel ␣ subunit inactivation are likely to be a primary function of the Kv␤1 and Kv␤3 gene products. Mice rendered deficient in Kv␤1 by gene targeting display reduced K ϩ current inactivation, frequency-dependent spike broadening, and slow after-hyperpolarization in hippocampal CA1 pyramidal neurons (8).
The Kv␤2 gene product, by far the most abundantly expressed Kv1-associated ␤ subunit protein (1,9), unlike Kv␤1 and Kv␤3, lacks the ball-and-chain domain and does not confer rapid inactivation to noninactivating Kv1 ␣ subunits. In heterologous expression systems it does, however, increase the rate of rapid inactivation of endogenously inactivating Kv1.4 currents (10). Kv␤2 also induces small shifts in the activation threshold of Kv1.5 currents (11) and alters inactivation rates when coassembled in combination with Kv␤1 subunits (10,12,13). In addition to the relatively restricted effects observed on K ϩ channel gating, Kv␤2 has also been reported to increase the amplitudes of Kv1.4 and Kv1.3 currents in Xenopus oocytes (4,10). In cultured mammalian cells the absence of Kv␤2 results in unglycosylated forms of Kv1.1 and Kv1.2 that are not efficiently transported to the membrane surface, leading Shi et al. (2) to propose that Kv␤2 facilitates the glycosylation of Kv1 ␣ subunits in the endoplasmic reticulum, thereby promoting the trafficking of Kv␣/␤ complexes from the endoplasmic reticulum to the surface cell membrane. Whether Kv␤2 is required for trafficking of these Kv␣ species in vivo is unknown.
Given their homology to other AKR 1 superfamily members, including the conservation of nucleotide-cofactor binding and catalytic residues, additional physiological functions for Kv␤2 have been proposed relating to potential oxidoreductive and/or catalytic roles (14,15). Interestingly, coexpression of Kv␤2 in cultured mammalian cells confers oxygen sensitivity to Kv4.2 but not Kv1 channels (16). Subsequently, the structural and potential functional similarities between Kv␤2 and other AKR superfamily members, including cofactor binding, was confirmed through x-ray crystallographic resolution (17). Kv␤2 binds NADPH with an affinity comparable with that found for other AKR enzymes but with only 10-fold lower affinity for NADH; NADPH and NADH bind more tightly than NADP ϩ and NAD ϩ , respectively, indicating that Kv␤2 would be more likely to function as a reductase than an oxidase (18); cofactor binding is not a prerequisite for multimer formation. It therefore remains possible that Kv␤2 is an oxidoreductive enzyme or acts as a metabolic sensor with regard to cofactor binding and interaction with at least some Kv␣ subunits. Despite much research into the expression of Kv␤2 in heterologous systems, little is known regarding the function of Kv␤2 in vivo. The human Kv␤2 locus (KCNAB2) maps to chromosome 1p36, where it has been implicated in complex syndromes that include seizures (19,20).
To investigate the possible functions of Kv␤2 in the intact mammalian nervous system, we used gene targeting in embry-onic stem cells to generate Kv␤2-null or Y90F point mutant (which should specifically inhibit typical AKR catalytic activity) mice. We conclude that Kv␤2 plays an important role in modulating excitability in vivo but that this role does not rely upon Kv␤2 enhancement of Kv1 ␣ subunit trafficking and/or targeting. We also find that the primary physiological role of Kv␤2 does not appear to take place through typical AKR-like oxidoreductive catalysis.

Construction of the Kv␤2-null Targeting Vector
The Kv␤2-null targeting vector was constructed from a genomic clone isolated from a 129/SvEv genomic library (Stratagene number 946305) as diagrammed in Fig. 1A. Exon/intron boundaries of mouse Kv␤2 as determined by our sequencing do not correspond with those previously reported for 129/SvJ mice (21), but do correspond with those reported for exons 5-9 of the human Kv␤2 gene locus (GenBank TM accession number NT_019265.6). The targeting vector contained a 2.5-kb 5Ј arm of homology (XhoI-NsiI fragment) and a 3-kb 3Ј arm of homology (SalI-BglII fragment). The internal NsiI-SalI fragment containing genomic DNA corresponding to exons 7-9 of human Kv␤2 was deleted such that any potential transcript would have a minimal deletion of 176 nucleotides of coding sequence; PGKneobpA (22) was inserted in place of the deleted fragment. MC1-thymidine kinase was attached at the 3Ј end for negative selection against nonhomologous recombinants.

Construction of the Y90F Point Mutant Knock-in Vector
Genomic DNA encoding exon 5 ( . . . agATG GCA GAG CAC CTA ATG ACC TTG GCC TAC GAT AAT GGC ATC AAC CTG TTC GAT ACG GCG GAG GTC TAC GCT GCT GGA Aagtacgtagt. . . . ) and residue Tyr 90 was isolated from the same genomic library. Using the AccI site (GTC TAC) overlapping Tyr 90 (TAC) and a SnaB1 site in the 3Ј intervening sequence (tactgta), double-stranded oligonucleotides containing three substitutions (underlined) were introduced into exon 5 (GTC TTC GCC GCG GGA Aagtacgta) mutating Tyr 90 to Phe 90 while removing the AccI site and introducing a SacII site. A 1.9-kb SalI/XhoI PGKneo fragment flanked by loxP sequences (gift of A. Nagy) was then cloned into the XhoI site ϳ100 nucleotides downstream of the 3Ј end of exon 5. The final targeting vector consisted of a contiguous 5.7-kb BamHI/BglII genomic fragment interrupted at the XhoI site with the floxed PGK-neo, with a 1.3-kb 5Ј arm of homology containing the point mutation and the 4.4-kb 3Ј arm of homology (see Fig. 1B).

Creation of Targeted Embryonic Stem Cells and Kv␤2-null and Y90F Mice
AB-1 ES cells (obtained from Allan Bradley) were electroporated with 10 g of linearized targeting vector and treated with G418 (350 g/ml) and FIAU (200 nM) for selection of doubly resistant colonies. Southern blot analyses were performed as described (23). For the ES cells transfected with the Kv␤2-null targeting vector, DNA was digested with BglII (for 5Ј probe) or HindIII (for neo probe). For the ES cells transfected with the Kv␤2-Y90F targeting vector, DNA was digested with BamHI (for the 3Ј probe) or AccI and AccI/SacII (for the neo probe). Excision of the PGKneo insert was induced through transient transfection of CMV-Cre recombinase (pBS185; Invitrogen), and recombinant clones were identified through Southern blot analysis utilizing the same 0.8-kb AvrII/SalI 3Ј probe by comparison of BamHI versus BamHI/SacII digestion as well as by PCR using primers corresponding to genomic DNA sequences 5Ј of exon 5 and 3Ј of the XhoI site in intron 5.
Targeted ES cell clones were injected into C57BL/6J (B6) blastocysts (University of Wisconsin Transgenic Core Facility) to generate chimeric founder mice. Founder mice were bred to inbred B6 mice to generate F1 heterozygotes, which were then bred together to generate mutant mice on a mixed B6ϫ129 genetic background. The Kv␤2-null and Y90F point mutant mice described in this report have been given the strain designations Kcnab2 tm1Mes and Kcnab2 tm2Mes , respectively.

Survival Analysis
All Kv␤2-null, Y90F, and heterozygous mice that were generated through the F2 and F3 generations (in mixed B6ϫ129 backgrounds) were included in the survival analysis. Mice that were alive as of an arbitrary end date (January 29, 2001) or were removed for various reasons (for example, use in experiments) were treated as censored data. Kaplan-Meier survival curves were generated for both groups and compared using the logrank test.

Myoclonus Score Following Cold Water Swim
Mice were placed in a tank of 17°C water for 2 min. The mice were then placed in a dry cage for observation by a scorer blinded to the genotype. The myoclonus score is the number of whole body jerks observed during the first 2 min post-swim. Kv␤2-null mice and wild type littermates were 42-58 days old, Kv␤2-Y90F and wild type littermates were 50 -53 days old, and Kv1.1-null mice (24) and wild type littermates were 30 -41 days old. All mice scored for myoclonus were on an inbred 129 background.

Immunofluorescence
Sciatic Nerve-Briefly, the sciatic nerves were immersion fixed with 4% paraformaldehyde in phosphate-buffered saline at 4°C and then teased into single fibers on glass slides and processed for Kv1.1, Kv1.2, and Kv␤2 immunofluoresence as previously described (27). Confocal images of immunofluorescence were captured using a Bio-Rad confocal microscope (MRC 1024). The published images were obtained by plane projection from 4 -10 consecutive z-section images with 0.2-m z-increments. During image capturing, the settings on the microscope (background, gain, and axis) were adjusted so that the peak readings at the juxtaparanode were always lower than 256 (the maximum value) to minimize saturation. Spatial distribution of the Kv1.1 and Kv1.2 immunofluorescence intensity along single fibers at the nodal region was quantitatively measured along a fixed length (60 m) using MetaMorph software. The line scan function was used with a setting of 6 ϫ 60 pixels (thickness ϫ length).
Cerebellum-All mice were anesthetized deeply with avertin and transcardially perfused with 60 ml of chilled fixative containing 15% picric acid, 4% paraformaldehyde in 0.1 M phosphate-buffered saline buffer. The brain was removed and immersed in the same fixative for another 3-4 h at 4°C. After several washes, the fixed brain was cyroprotected in 30% sucrose solution in 0.1 M phosphate-buffered saline overnight. Frozen sagittal sections of the cerebellum were cut at 50 m and permeabilized in blocking solution for 2 h with 5-10% goat serum, 0.4% Triton X-100, and 0.02% sodium azide in 0.1 M phosphate-buffered saline. The subsequent staining and imaging were similar to that described above for sciatic nerves.

Western Blot Analysis of Kv␣ Subunits
Age-matched littermate or C57BL/6J mice were used as wild type controls. For Western blot analysis of Kv␣ subunits, whole brains were removed, washed several times with ice-cold lysis buffer (150 mM NaCl, 10 mM Tris, pH 8.0, 1% Triton X-100, 1 mM EGTA, 1 mM EDTA) and then homogenized with 10 -15 strokes in a glass homogenizer in buffer containing 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 2 g/ml aprotinin, 2 g/ml antipain, and 10 g of benzamide as protease inhibitors. The extracts were sonicated for 1 min and then spun in a microcentrifuge for 4 min at 4°C. The supernatant was quickly mixed with 5ϫ sample buffer (250 mM Tris, pH 6.8, 500 mM dithiothreitol, 10% (w/v) SDS, 0.5%, bromphenol blue, 50%, glycerol) in a 4:1 ratio and boiled for 2 min. The protein concentration in each sample was quantified by a Bradford assay (Bio-Rad), and the volume of each sample was adjusted so that all were at 20 mg/ml. The proteins were separated on a 9% SDS-polyacrylamide gel and then transferred to nitrocellulose membranes (Bio-Rad; 0.45 m). The membranes were blocked in TBS-Tween (20 mM Tris, pH 7.6, 137 mM NaCl, 0.1% (v/v) Tween 20) and then incubated in primary antibodies (monoclonal anti-Kv1.1, Kv1.2, and Kv1.4, used at 1:1000, 1:500, and 1:1000, respectively). The membranes were then washed four times for 5 min before incubation with horseradish peroxidase-conjugated secondary antibodies (Pierce; 1:3000 dilution). After washing, the membranes were developed with enhanced chemoluminescence (Pierce) for 1 min and then exposed to film. To verify equal loading of lanes, the membranes were stained following ECL with Ponceau S for 5 min, followed by fixation with 5% glacial acetic acid.
For analysis of glycosylation status of Kv␣ subunits, the brain extracts were incubated overnight at 37°C with either N-glycosidase F (PNGase F; 20 units/ml) or endoglycosidase H (EndoH; 0.5 unit/ml), both purchased from Roche Molecular Biochemicals. For PNGase treatment, the brain extract was first diluted 1:10 with lysis buffer without SDS.
For semi-quantitative analysis of Kv␣ subunit expression, binding of primary antibodies were detected using 125 I-labeled protein A. After transfer of proteins, the nitrocellulose membranes were blocked in TBS-Tween with 5% dried milk, washed, incubated with primary antibodies as described above, and then washed again. The membrane was next incubated with goat anti-rabbit or rabbit anti-mouse secondary antibodies, depending on the species of the primary antibody. After another wash, 0.5 Ci of 125 I-conjugated protein A was diluted in 10 ml of block solution and incubated with the membrane for 1 h at room temperature. Finally the membrane was washed five times for 5 min, air-dried, and then wrapped in plastic and mounted in a Phosphor-Imager cassette. The upper and lower bands, which correspond to the mature and core glycosylated forms (see below), were analyzed together with background subtraction. The ratio of upper band to lower band in each sample was calculated directly.

Western Blot Analysis of Kv␤ Subunits
For Western blot analysis of Kv␤ subunits, whole brains were homogenized in 0.3 ml of 10 mM sodium phosphate, pH 7.4, 10 mM sodium fluoride homogenization buffer, containing 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 2 g/ml aprotinin, and 1 g/ml pepstatin. The homogenate was centrifuged at 3000 ϫ g for 10 min to remove nuclei and cellular debris, and the supernatant was used for Western blot analysis. Protein content was assayed by the Bradford assay. The samples of supernatant (10 l, containing 50 g of protein) were boiled in SDS sample buffer containing 63 mM Tris-HCl, pH 6.8, 1% SDS, 1% ␤-mercaptoethanol, 10% glycerol, and 0.001% bromphenol blue for 10 min. The proteins were then separated on a 15% polyacrylamide gel and transferred onto polyvinylidene difluoride-plus membrane. The membranes were blocked in 3% nonfat milk in Tris-buffered saline (Tris-HCl, pH 7.4, 150 mM NaCl) overnight at 4°C and probed with monoclonal anti-Kv␤1 (K9/40.1) or anti-Kv␤2 (K17/70) antibodies (25) at room temperature, followed by horseradish peroxidase-conjugated rabbit anti-mouse (Pierce) and ECL detection.

Electrophysiology
Kv1.4 cRNA (30 pg) was coinjected with or without just-saturating amounts of Kv␤2 or Kv␤2-Y90F cRNA (60 pg) into Xenopus oocytes. Three days later oocyte membranes were depolarized to ϩ50 mV from a holding potential of Ϫ90 mV using standard two-microelectrode recording techniques as described elsewhere (10).

RESULTS
Generation of Kv␤2-null Mice-Kv␤2-null mice were generated by gene targeting in embryonic stem cells (Fig. 1A). The targeting vector was designed to introduce a deletion toward the beginning of the constitutive "core" region of the protein that includes essential ␤-sheet structural elements (14,17) such that any ␤2 gene products would be structurally compromised, unlikely to fold, and rapidly degraded. Following electroporation of the targeting vector into ES cells, 80 clones were isolated that were doubly resistant to G418 and FIAU. Southern blot analysis revealed that 11 of these contained the disrupted Kcnab2 locus (data not shown). One clone (K59) was used to generate chimeric founder mice and transmission of the mutant allele to heterozygous offspring, which were interbred to produce homozygous mutant mice. These crosses yielded the expected Mendelian ratios of wild type, heterozygous, and homozygous mutant mice (ϩ/ϩ ϭ 32, ϩ/Ϫ ϭ 78, Ϫ/Ϫ ϭ 32; 2 ϭ 1.38, df ϭ 2, p ϭ 0.50). The data presented below demonstrated that the mutant mice express no detectable Kv␤2 protein in the peripheral or central nervous systems.
Kv␤2-null mice also exhibit a cold swim-induced tremor. Mice were forced to swim in a tank of water for 2 min at 17°C and then observed on a dry platform, where they displayed whole body tremors indicative of hyperexcitability (Table I). This phenotype was not observed after swimming in 37°C water and was never observed in age-matched wild type controls. Interestingly, similar but somewhat stronger phenotypes were observed in Kv1.1-null (␣ subunit) mutants with respect to cold swim-induced tremors, reduced life span, and spontaneous seizures (24,27). The phenotypic effects are therefore consistent with the possibility that Kv␤2-null mice have reduced expression of Kv1.1 or other Kv1 proteins caused by abnormal biogenesis or trafficking, as has been observed in heterologous expression systems (2). If so, one would expect that the absence of Kv␤2 would result in significant reductions in Kv1 ␣ subunit expression levels, given that Kv1.1 heterozygotes display no tremors following exposure to cold water and exhibit normal life spans (data not shown), despite substantial reductions in mRNA (24) and protein (see Fig. 7).
Kv1.1/Kv1.2 Trafficking and Biogenesis in Kv␤2-null Mutants-If Kv␤2 proteins play an important role in the early biogenesis and trafficking of Kv1 channels in vivo, in Kv␤2-null mice the channels should become trapped in the endoplasmic reticulum and exhibit a lack of Golgi-processed glycosylation (2). To assess the processing of Kv1 ␣ subunits, we analyzed glycosylation patterns of Kv1.1 and Kv1.2 in Western blots of brains from wild type and Kv␤2-null mice. Both the wild type and Kv␤2-null mice show high (ϳ88 kDa) and low (ϳ60 kDa) molecular mass species for Kv1.2 (Fig. 3, left lanes) and Kv1.1 (data not shown).
The two bands observed for Kv1.2 are thought to represent different glycosylated forms; the low and high molecular bands correspond to the core-glycosylated and mature-glycosylated forms of the ␣ subunit, respectively (2,28). To verify this classification, we tested the susceptibility of these bands to two widely used glycosidases: EndoH and PNGase F. EndoH preferentially cuts the high mannose type glycanes from asparagine found in core-glycosylated forms, whereas PNGase cleaves all asparagine-linked glycanes. With EndoH, the high molecular mass band was unaffected, whereas the lower molecular mass band was reduced ϳ3 kDa (Fig. 3). These results are consistent with previous reports that core glycosylation leads to 2-3-kDa increases in molecular mass (28). In contrast, PNGase treatment reduced the high molecular mass band by ϳ20 kDa, consistent with it being the mature glycosylated form.
In Kv␤2-null mice the glycosylation patterns of Kv1.2 subunits in Kv␤2-null mice appeared comparable in size and relative proportions to those from wild type mice, and no bands were found at the size predicted for unglycosylated Kv1. Quantitative Analysis of Western Blots-To quantify the proportion of Kv1␣ subunits that are glycosylated in the mutant mice, we performed two types of quantitative Western blot analysis from whole brain extracts. Fig. 4A represents serial dilutions from brains of a representative pair of wild type and Kv␤2-null mice at p45, probed for Kv1.2 and detected by chemiluminescence, indicating little or no difference in the intensity of the high molecular mass band. Similar results were obtained in blots probed for Kv1.1 (data not shown). In a second type of analysis, we probed blots for Kv1.2 using 125 I-labeled protein A with PhosphorImager detection and compared the ratio of the high versus low molecular mass bands in each sample from four independent sets of age-matched wild type and Kv␤2-null mice (Fig. 4B). Because each ratio (high versus low molecular mass bands) is calculated against an internal control (the low molecular mass band), these values would not be subject to minor differences in protein loading between lanes. Although the ratios for mutants ranged from 72 to 104% that of their agematched wild type control in these four samples, the differences were not statistically significant (paired t test, t ϭ 1.933, df ϭ 3, p ϭ 0.149). These results confirmed that there is little or no change in the proportion of Kv␣ subunits that achieve mature glycosylation in the absence of Kv␤2. Introduction of the Y90F mutation into exon 5 also results in conversion of an AccI site to a SacII site. Southern blot analysis using the 0.8-kb AvrII-SalI fragment (shown as probe c in a shaded box) and BamHI digestion gives an 11.3-kb band for the wild type allele and a 13.1-kb band for the initial targeted allele. Recombination at the 5Ј end was verified using the PstI-NotI fragment from neo as a probe and AccI for digestion, and inclusion of the Y90F mutation was verified by an AccI-SacII digestion. Excision of the PGK-neo to generate the final recombinant allele was verified using the 3Ј probe with BamHI versus BamHI-SacII digestion. The residual loxP site is shown as an open triangle in intron 5. A, AccI; Av, AvrII; B, BamHI; Bg, BglII; S, SacII; Sl, SalI; X, XhoI. significant phenotypic consequences for the animal. To verify the absence of Kv␤2 protein in the null animals, we evaluated central and peripheral nervous system tissues from Kv␤2-null mice using a monoclonal antibody directed against the N-terminal region of the protein. We focused on two regions of the nervous system where colocalization of Kv1.1/Kv1.2/Kv␤2 has been extensively studied: the terminals of cerebellar basket cells and the juxtaparanodes of myelinated nerves (29). As shown in Fig. 5, immunofluorescent staining for Kv␤2 in either cerebellum or myelinated nerve fibers of control mice reliably shows prominent labeling of basket cell terminals or axons juxtaparanodes, respectively. However, only background fluorescence is detected in the Kv␤2-null mice. Moreover, neither full-length nor truncated Kv␤2 protein products were observed in Western blot analyses of whole brain lysates from the null mice (data not shown).
We subsequently examined Kv1␣ expression and distribution in cerebellar cortex by immunofluorescent staining. Consistent with our results from Western blot analysis of whole brain lysates, which showed apparently normal glycosylation, Kv1.2 in the cerebellar cortex was properly concentrated, and at approximately comparable levels, in the basket cell terminals of Kv␤2-null mice (Fig. 6). Similar results were obtained for Kv1.1 (data not shown). Furthermore, no labeling of basket cell bodies was observed in the molecular layer as might have been expected if Kv1␣ subunits became trapped within the endoplasmic reticulum.
We further investigated the expression and targeting of Kv1␣ subunits in peripheral myelinated fibers by comparing the axial distribution of Kv1.1 along the paranode-node-paranode region of single fibers using quantitative fluorescence. Initially, we assayed the sensitivity of our method by analyzing paranodes from Kv1.1 heterozygous mice. In such mice, Northern blot analysis indicates a reduction of brain mRNA to 54% of wild type levels (24), and electrophysiological analysis indicates a reduction of Kv1.1 protein level in sciatic nerves (30). Fig. 7A represents fluorescence profiles, with valleys centered at the node (where Kv1.1 is absent) flanked by two peaks at the juxtaparanodes (where Kv1.1 density is highest), for the spatial distribution of Kv1.1 in single fibers from Kv1.1 heterozygote and wild type fibers. The averaged peak juxtaparanodal intensity from three Kv1.1 heterozygotes is about 67% that of wild type, indicating that this method is capable of detecting reduction in Kv1 protein levels at the juxtaparanodes.
Similar comparison of Kv1.2 (Fig. 7B) and Kv1.1 (Fig. 7C) immunoreactivity in sciatic nerves from wild type and Kv␤2null mice indicated that the expression and distribution of these ␣ subunits were not appreciably affected by the absence of Kv␤2. For Kv1.1 and Kv1.2, the averaged ratio of the peak paranodal intensity of Kv␤2-null mice to wild type was 1.02 (three pairs of mice) and 0.954 (two pairs of mice), respectively. Taken together, the glycosylation, trafficking, surface expression, and clustering of Kv1 channels in vivo was not significantly affected by the absence of the Kv␤2 gene product.
Although Kv␤2 is by far the most abundantly expressed Kv1 ␤ subunit in brain (1,9), we examined whether upregulation of related ␤ subunits such as Kv␤1 might compensate for the loss of Kv␤2 through Western blot analysis of the Kv␤1protein on whole brain extracts from wild type and Kv␤2-null mice. No evident up-regulation of Kv␤1 was observed (data not shown).
Electrophysiological Effects of Native and Mutant Y90F Kv␤2 in Xenopus Oocytes-To investigate the potential enzymatic role of Kv␤2 in vivo, we sought to create a mutation of the protein that would abolish enzymatic activity without altering the stability or potential nucleotide binding properties of the protein. Mutation of the Tyr 90 residue, which is homologous to the tyrosine of other AKR family members where it plays a direct role in catalysis (31), to Phe 90 was introduced into the Both high (ϳ88 kDa) and low (ϳ60 kDa) molecular mass bands are detected for Kv1.2 in wild type and in Kv␤2-null brains (left lanes), with a representative pair of animals illustrated that were 20 -21 days old. Upon digestion with EndoH, the high molecular mass band was unaffected, whereas the lower molecular mass band was reduced ϳ3 kDa, suggesting that the low molecular mass band represents the core-glycosylated form. In contrast, PNGase treatment reduced the high molecular mass band by ϳ20 kDa, consistent with this being the mature glycosylated form. PNGase digestion also resulted in a shift of the lower band to a position similar to that in the EndoH digest. Molecular mass marker sizes (in kDa) are indicated. The PN-Gase lanes were exposed longer than the other lanes.  (24), is seen in Kv␤2 nulls but not Y90F mutant mice. Mice were placed in a tank of 17°C water for 2 min, and then a myoclonus score was determined by counting the number of whole body jerks that occurred during the first 2 min following removal from the water. The values shown indicate the score Ϯ S.E. Body weights (g Ϯ S.E.) were comparable for each set of Kv␤ mutants and their controls, although Kv1.1 mutants were slightly smaller than their wild type controls (for Kv1.1, wild type ϭ 15.0 Ϯ 0.5, mutant ϭ 12.1 Ϯ 0.7; for Kv␤2, wild type ϭ 20.0 Ϯ 0.7, mutant ϭ 21.2 Ϯ 0.9; for Y90F, wild type ϭ 16.9 Ϯ 0.5, mutant ϭ 18.1 Ϯ 0.9). human Kv␤2 cDNA. Kv␤2-Y90F was subsequently coexpressed with human Kv1.4 channels in Xenopus oocytes as an assay for its stability and ability to interact with Kv1 channels (10). Equivalent mutations in several other AKR members to any amino acid residue other than histidine completely abolish enzymatic activity while having little effect on nucleotide binding or stability (31,32). Both native and mutant Y90F ␤2 subunit proteins accelerate the inactivation and increase the current amplitudes of Kv1.4 currents (Fig. 8) similar to that reported previously (10). This result indicates that the wild type ␤2 protein increases the rate of Kv1.4 inactivation and current amplitudes, either through nonenzymatic or atypical AKR enzymatic processes. It also suggests that the Y90F ␤2 protein is stable. Although the mutation would be expected to inhibit any inherent AKR-like enzymatic function, the protein should be able to carry out any other physiological functions including atypical AKR activity (see "Discussion") and cofactor binding.
Generation of Kv␤2 Y90F Mice-To test the role of putative oxidoreductive catalysis by the Kv␤2 subunit in vivo, mice containing the Y90F point mutation of the Kcnab2 gene were generated in two steps. Following electroporation of the target-ing vector (Fig. 1B) into ES cells, 90 clones were isolated that were doubly resistant to G418 and FIAU. Seven of these clones contained the mutated Kcnab2 locus (data not shown). One clone (K91) was subsequently expanded and transiently transfected with a Cre recombinase expression vector; subsequent Southern blot analysis identified 17 of 96 subclones having undergone excision and loss of PGKneo (data not shown). One subclone (K113) was chosen to generate chimeric founder mice, which subsequently demonstrated germline transmission of the mutant allele. Mating F1 heterozygotes yielded the expected Mendelian ratios of wild type, heterozygous, and homozygous mutant mice (ϩ/ϩ ϭ 22, Y90F/ϩ ϭ 47, Y90F/Y90F ϭ 28; 2 ϭ 0.84, df ϭ 2, p ϭ 0.66), as expected given the lack of embryonic lethality in the Kv␤2-null mutation. Expression of Y90F Kv␤2 RNA transcripts was verified using reverse transcription-PCR in combination with AccI and SacII digests (data not shown).
Y90F Kv␤2 Mice Do Not Show Phenotypic Alterations Observed in Kv␤2-null Mutants-The Y90F mutant mice were not observed to undergo spontaneous seizures and were similar to wild type mice in both the swim test and life span analysis (Table I and Fig. 2; for life span analysis, Y90F/ϩ n ϭ 21 and A, Western blot analysis was carried out on brain extracts from wild type and Kv␤2-null littermates, with a representative pair shown here at p45. The samples were adjusted so that protein concentrations were the same and then subjected to serial dilutions before loading on the gel. After development with ECL and exposure to film, the membrane was stained with Ponceau S to reveal total protein and comparable loading between wild type and mutant lanes. 16 g of total protein was loaded in the 1:1 lane. B, Western blot analysis using 125 I detection was carried out on brain extracts from four pairs of age-matched wild type and Kv␤2-null mice. The nitrocellulose membrane was coincubated with monoclonal antibodies to both Kv1.2 and the 145-kDa neurofilament protein (as an internal control for loading), then incubated with a rabbit anti-mouse antiserum, and then finally incubated with 125 I-labeled protein A. After detection by PhosphorImager, the image was analyzed by Quantity-One (Bio-Rad). The ratio of the high molecular mass to low molecular mass bands for Kv1.2 were calculated for each mouse (after background subtraction), and then the ratio of mutant to wild type for each pair was calculated and subjected to paired sample statistical analysis as described under "Results." Approximately 50 g of protein was loaded in each lane.
Y90F/Y90F n ϭ 77). These results strongly indicate that mutant phenotypes associated with the Kv␤2-null mutant mice involve loss of a nonenzymatic function or that Kv␤2 exhibits atypical AKR catalytic activity. DISCUSSION In this paper, we describe the generation of Kv␤2-null and Kv␤2-Y90F mutant mice to examine the in vivo function of Kv␤2, a major auxiliary subunit of Kv1␣ channels in the mammalian nervous system. Kv␤2-null mice display cold swim-induced tremors, occasional seizures, and reduced life spans similar to that observed in Kv1.1-null mice. We therefore examined whether the absence of Kv␤2 might result in a failure of Kv1␣ subunit surface expression in vivo. We turned our attention to the cerebellar basket cell terminals and the paranodal regions of myelinated fibers, both of which are normally characterized by a striking colocalization of Kv1.1, Kv1.2, and Kv␤2 gene products but relatively free of Kv␤1 (33,34). Thus, disruption of a putative chaperone-like role of Kv␤2 might be expected to produce an obvious impact at these two regions.
Kv1 Channel Surface Expression-The ability of Kv␤1 subunits to confer enhanced surface expression of Kv1␣ K ϩ channels in cultured cell lines was first demonstrated by Shi et al. (2). In that study, Western blot analysis indicated that expression of Kv1.2 in COS cells resulted in production of 63-kDa (unglycosylated) and 66-kDa (glycosylated) bands of FIG. 5. Absence of Kv␤2 proteins in Kv␤2-null cerebellum and sciatic nerves. Cerebellar slices (top row, 3-month-old mice) or teased sciatic nerve fibers (bottom row, 4-month-old mice) were stained with monoclonal anti-Kv␤2 antibody in wild type (left panels) and Kv␤2-null (right panels) mice. Normal localization is seen in the wild type controls, but no detectable Kv␤2 protein is seen in the null mutants. Scale bar, 15 m. roughly equal intensity, with a significant amount of Kv1.2 immunoreactivity retained in an intracellular compartment (likely to be endoplasmic reticulum). When coexpressed with Kv␤2, most of the Kv1.2 was in the glycosylated 66-kDa band, and stronger Kv1.2 immunoreactivity was observed at the cell surface. Shi et al. (2) therefore proposed that the primary physiological role of Kv␤2 is to facilitate the folding, glycosylation, and trafficking of Kv1 proteins to the cell membrane. More recently, Campomanes et al. (35) found that Kv␤ subunits promoted axonal targeting of transfected Kv1.2 in primary neuronal cultures.
In contrast to the studies carried out in cultured cells, the absence of Kv␤2 in vivo does not lead to a detectable retention of Kv1.1/Kv1.2 immunoreactivity in the cell bodies of the cerebellar basket cells. Furthermore, our immunofluorescence studies show that both the clusters of Kv1.1/Kv1.2 proteins at the basket cell terminal and the juxtaparanodal regions of myelinated axons are not affected in the Kv␤2-null mutants. In addition, our Western blot analysis failed to reveal accumulation of unglycosylated species for Kv1.1 and Kv1.2 in Kv␤2-null mutants, suggesting that the Kv1␣ subunits are glycosylated efficiently in vivo without Kv␤2. Our results are most consistent with the data of Nagaya and Papazian (36), who found that the ␤2 subunit did not increase the rate or extent of Shaker protein maturation in transfected mammalian cells.
We considered two ways in which a chaperone-like role of Kv␤2 might have eluded our detection. First, other Kv␤ subunits such as Kv␤1 may compensate for the loss of Kv␤2. However, this appears unlikely because Kv␤1 is normally absent in the basket cell terminals and the juxtaparanodal regions of myelinated fibers, and we detected no significant upregulation of Kv␤1 in Western blot analysis of whole brain lysates from Kv␤2-null mice. Although a third Kv␤ subunit has been identified in rat brain, little is known regarding its cell specificity, level of expression, and ability to associate with or assist in trafficking of Kv1␣ subunits in vivo (6). Second, our biogenesis analysis was performed on mice between the ages of P30 -P40, and given the reduced life span of mutant mice older than 5 months, we cannot exclude the possibility that a significant biogenesis defect emerges in older mice.
K ϩ Channel Subunit Composition-Most individual neurons expressing Kv1 channels appear to express multiple family members, and some Kv1 ␣ subunits may facilitate efficient surface expression of heteromultimeric Kv1 channels in the absence of Kv␤2 (37). For example, homotetramers of Kv1.2 and Kv1.4, but not Kv1.1, are expressed efficiently on the surface of mammalian cell lines; Kv1.1/Kv1.2 or Kv1.1/Kv1.4 heterotetramers are expressed more efficiently than Kv1.1 is by itself. Intrinsic differences in the trafficking of specific Kv1 ␣ subunits may be regulated to some degree by Kv␤2, and its absence might therefore lead to changes in the composition of native multimeric Kv1 channels and contribute to the phenotypic effects observed in Kv␤2-null mice.
Loss of Kv␤2 might also affect the formation of channels between Kv1 ␣ subunits and those from different families, such as eag (38) or Kv4 (3). Alternatively, the absence of Kv␤2 may change the ability of Kv4 to interact with other known binding partners, including the KChIP family of proteins (39). Studies of K ϩ currents in the Drosophila mutant hyperkinetic (which lack K ϩ channel ␤ subunits) have shown altered sensitivity to K ϩ channel blockers (40), consistent with a change in Kv␣ subunit composition of native channels in the mutant. Preliminary studies of the neuromuscular junction in our Kv␤2-null mutant mice revealed a dramatically heightened sensitivity to TEA 2 ; whether this altered pharmacology reflects a change in the Kv␣ composition of native K ϩ channels in the Kv␤2-null mice remains to be established.
K ϩ Channel Kinetics-Another possible mechanism that could contribute to the phenotype of Kv␤2-null mice is alteration of K ϩ channel kinetics caused by modulation of K ϩ channel inactivation. Despite the absence of the ball-and-chain motif to rapidly inactivate ␣ subunits (the N terminus of Kv␤2 is shorter than those of Kv␤1 and Kv␤3), Kv␤2 has been shown to modulate ␣ subunit kinetics directly, as has also been shown for Kv␤ subunits in Drosophila (15). Kv␤2 accelerates Kv1.4 inactivation (10, 41), accelerates Kv1.5 activation, and shifts the Kv1.1 and Kv1.5 activation threshold (11). Kv␤2 could also modulate ␣ subunit inactivation indirectly by competing with other ␤ subunits that bind to ␣ subunits. For instance, Kv␤2 alters the kinetic effects of Kv1 channel inactivation induced by Kv␤1 in coexpression experiments (10,12,13). Deletion of Kv␤2 might favor saturation of Kv␤1 binding to some individual ␣ subunits, resulting in faster inactivation of a proportion of K ϩ currents. Faster inactivation could lead to prolongation of the action potential, faster firing frequencies, and hyperexcitability.
Do Kv␤ Subunits Have an Enzymatic Function?-Kv1 ␤ subunits are members of an aldo-ketoreductase enzyme superfamily that catalyze the oxidation/reduction of a wide range of physiologic substrates (14,15,31,42), and crystallographic and binding studies reveal a tightly bound NADPH cofactor (17,18). Although sequence homology, structural resolution, and critical residue conservation have all strongly suggested Kv␤2 as a potential catalytic AKR, no specific enzyme activity has been directly identified for the Kv␤2 protein to date. The enzymatic activity of Kv␤2, if it exists, may be difficult to investigate for a number of reasons, including the identity of a potential substrate and the hypothesis that the catalytic activity of Kv␤2 may be induced by the activity of functional (and therefore intact) Kv ␣ subunits (14,17). It is difficult if not impossible to assay a large number of potential substrates biochemically given the potential necessity for reconstitution of 2 L. Zhou and S.-Y. Chiu, unpublished observations. FIG. 8. Electrophysiological properties of coexpressed Kv1.4 and wild type or Kv␤2-Y90F in Xenopus oocytes. A, single exponential inactivation time constants were obtained from the decay of K ϩ currents obtained from 200-ms depolarizing pulses to ϩ50 mV from a holding potential of Ϫ90 mV. B, current amplitudes were obtained from the peak currents obtained using an identical pulse protocol.
Kv␣/␤ complexes with active electrophysiological stimulation. We therefore specifically chose to address this question using genetic methods and focusing on what is widely accepted as the essential residue for AKRs, Tyr 90 . The straightforward question being asked is whether Y90F mice show any overt phenotype, and if so whether it is similar or identical to that observed for the Kv␤2-null mice? The unequivocal answer to these questions is no.
Several recent publications provide indirect evidence for the potential enzymatic function of Kv␤ proteins through modification of Kv1 subunit properties in heterologous expression systems (35,43,44). These studies have specifically limited their analysis to the effects of Kv␤2 catalysis and/or substrates on Kv1 ␣ channels heterologously coexpressed with Kv␤, although there is no a priori reason to suggest that the actual substrate(s) must act upon the Kv1 channels. In contrast, phenotypes potentially detectable in Y90F mice would not be limited in this way and could result from deficient catalysis of any cellular substrate(s). Although more sophisticated testing will need to be performed before one can exclude an enzymatic role for the Kv␤2 protein, given the discrepancy in phenotypes between the Kv␤2-null and Y90F mutant mice, it appears that catalytic activity is not the primary physiological role of Kv␤2 or that Kv␤2 has atypical AKR enzymatic activity.
Mutating the catalytic tyrosine of other AKR enzymes consistently reduces or abolishes their catalytic activity. However, one of these enzymes, ␣-hydroxysteroid dehydrogenase, can still catalyze the reduction of quinone substrates with a mutation at the tyrosine residue but not at detectable levels with mutations of the catalytic lysine and only at minimal levels with mutations in the catalytic aspartate residues (42). Similarly, frog lens -crystallin, a naturally occurring AKR "mutant" that does not contain a catalytic tyrosine and whose main role is structural, also shows catalytic activity toward the nonphysiological quinone substrates (45). Whether mutations in Kv␤2 other than Y90F at putative catalytic or cofactor binding sites (35,43,44) will reveal phenotypic effects in the mouse system remains to be established.
In addition, conformational changes of Kv␤2 upon binding to substrate (ketone-or aldehyde-containing organic molecules) and/or cofactor might result in changes that involve kinetics, amplitude, or conformation of at least some K ϩ channel ␣ subunits. In this way, Kv␤2 could function as a local redox sensor that links K ϩ channel function to cellular metabolism. Local redox environment can be modulated by many different factors, including metabolic load, free radicals such as superoxide ion O 2 . or NO generated by synaptic activity (46,47). It has recently been reported that some K ϩ channels play a role in oxygen sensing (48), and Kv␤2 has been suggested to confer oxygen sensitivity to certain Kv4 ␣ subunits (16). Consistent with such a role, deletion of Kv␤2 may alter the sensitivity of some K ϩ channel activity to oxidative stress.
In conclusion, genetic deletion of Kv␤2 in mice leads to cold swim-induced tremors, sporadic seizures, and a reduced life span. The existence of a behavioral phenotype suggests that the function of Kv␤2 in vivo is not redundant and that compensation from other Kv␤ genes, if present, is limited and cannot provide full functional restoration. No apparent defects could be detected in biogenesis and trafficking of Kv1.1/Kv1.2 in the Kv␤2-null mice in the cerebellar basket cells and the myelinated fibers, nor were any alterations detected in glycosylation of Kv1 channels in whole brain lysates. These results suggest that the abnormal excitability phenotype in Kv␤2-null mice results from alterations in K ϩ channel functions that are still not well characterized and differ from the chaperone-like effects reported in heterologous systems (2, 10, 28, 44, 49). However, alterations in the composition and/or physiological properties of native Kv channel ␣/␤ complexes may contribute to the observed phenotype of Kv␤2-null mice. In addition, the lack of similar phenotypic effects in Kv␤2-Y90F mice indicate that typical AKR oxidoreductive catalytic activity is unlikely to be the primary physiological role of Kv␤2 gene products.