Selective Interaction of Voltage-gated K (cid:49) Channel (cid:98) -Subunits with (cid:97) -Subunits*

To begin to study the molecular bases that determine the selective interaction of the (cid:98) -subunits of voltage-gated K (cid:49) channels with (cid:97) -subunits observed in situ , we have expressed these polypeptides in transfected mammalian cells. Analysis of the specificity of (cid:97) / (cid:98) subunit interaction indicates that both the Kv (cid:98) 1 and Kv (cid:98) 2 (cid:98) -subunits display robust and selective interaction with the five members of the Shaker -related (Kv1) (cid:97) -subunit subfamily tested. The interaction of these (cid:98) -subunits with Kv1 (cid:97) -subunits does not require the (cid:98) -subunit N-terminal domains. Thus, the previously observed failure of N-terminal mutants of Kv (cid:98) 1 to modulate inactivation kinetics of Kv1 family members is not simply due to a lack of subunit interaction. Interaction of these (cid:98) -subunits with members of two other subfamilies ( Shab - and Shaw -related) could not be detected. Somewhat surprisingly, a member of the Shal -related subfamily was found to interact with (cid:98) -subunits; however, this interaction had biochemical characteristics distinct from the (cid:98) -subunit interaction with Kv1 family members. In all cases, Kv (cid:98) 1 and Kv (cid:98) 2 exhibited indistinguishable (cid:97) -subunit selectivity. These studies point to a selective interaction between K

Voltage-dependent K ϩ channels are fundamental and diverse components of neuronal activity. Molecular cloning studies have identified over a dozen distinct K ϩ channel genes and shown that the encoded pore-forming ␣-subunits are members of a large, multigene superfamily that includes Na ϩ and Ca 2ϩ channel ␣-subunits (1). Although expression of these individual ␣-subunits alone is sufficient to generate voltage-gated channels exhibiting many features of the corresponding channels in situ, studies on native Na ϩ and Ca 2ϩ channels in neurons and other excitable cells have confirmed the existence of auxiliary polypeptides in tight association with ␣-subunits (2). Cloning of these auxiliary subunits and their subsequent co-expression with ␣-subunits has shown that the expression level, gating, and conductance properties of expressed channels are profoundly influenced by the pres-ence of auxiliary subunits (2).
Recently, it has been discovered that K ϩ channels also have auxiliary (␤) subunits. A cDNA encoding a ␤-subunit copurifying with the bovine brain DTX acceptor complex was recently isolated (3). Subsequently, cDNAs encoding three highly related yet distinct ␤-subunit isoforms were isolated from rat brain (Kv␤1 and Kv␤2, Ref. 4) and from ferret (Kv␤3, Ref. 5) and human (hKv␤3, Refs. 6 and 7) heart. Although dissimilar in their primary structures, ␤-subunits of K ϩ and Ca 2ϩ channels exhibit general structural similarity in that they are basic (pI Ϸ 9.5), hydrophilic, and presumably peripheral membrane proteins present at the cytoplasmic face of the plasma membrane (2).
Co-expression of Kv␤1 was found to greatly accelerate the rate of inactivation of K ϩ currents expressed from the Kv1.1 or Kv1.4 ␣-subunit cDNAs in Xenopus oocytes (4). Kv␤3, which is an alternatively spliced product of the Kv␤1 gene, accelerates the rate of inactivation of K ϩ currents expressed from Kv1.4 or Kv1.5 but not from Kv1.1, Kv1.2, or Kv2.1 cDNAs (5)(6)(7). These results suggest that ␤-subunit modulation of ␣-subunit gating can contribute additional functional diversity to K ϩ channels in excitable cells. Surprisingly, coexpression of the highly related Kv␤2 had no effect on inactivation, apparently due to the lack of the N-terminal "ball" domain present in Kv␤1 that is both necessary and sufficient for the observed modulation of inactivation (4). However, from the published electrophysiological analysis of ␣/␤-subunit interaction presented, it was also possible that the lack of observed Kv␤2 effects was simply due to a lack of Kv␤2 interaction with the co-expressed ␣-subunits.
We previously used an antibody raised against the C terminus of the bovine ␤-subunit, predicted to recognize both Kv␤1 and Kv␤2 in rat brain, to investigate the expression of these ␤-subunits in situ (8). A major 38-kDa polypeptide and a minor 41-kDa polypeptide were detected in rat brain membrane fractions by immunoblot analysis. These two bands correspond closely to the predicted sizes of Kv␤2 and Kv␤1, respectively. Immunoprecipitation experiments showed that the major 38-kDa polypeptide is associated and colocalizes with Kv1.2 and Kv1.4, but not Kv2.1, in rat brain (8), suggesting the selective interaction of K ϩ channel ␣and ␤-subunits.
Cloning of Kv␤1 and Kv␤2 cDNAs-We initially cloned a 475 bp fragment of Kv␤1 cDNA by reverse transcriptase-PCR. Total adult rat brain RNA (1 g) was reverse transcribed, and the resultant cDNA was subjected to PCR. Oligonucleotide primers B1-5-1 (21-mer, 5Ј-GGAAT-TCTGAGAGGACCTTGC-3Ј) and B1-3-1 (20-mer, 5Ј-TTCTTCCAT-GGGGGTGTTGC-3Ј) were used for 20 rounds of PCR. Amplified product was fractionated by an agarose gel, and a specific product of predicted size (642 bp) was isolated and subjected to 15 additional rounds of amplification using the same primers. The resultant product was isolated from an agarose gel, digested with EcoRI and HindIII, and cloned into Bluescript SK ϩ . Its identity as a fragment of Kv␤1 cDNA was verified by sequencing. The clone was then used as a probe to screen a rat brain cDNA library (in ZAPII, kindly provided by Dr. T. Snutch (University of British Columbia)). From a total of 3.5 ϫ 10 5 plaques, 9 Kv␤1 clones were obtained. The identity of these clones was confirmed by restriction mapping and partial sequencing. One of the cDNA clones, pKB16, which contains the full-length Kv␤1 coding region, 0.3 kbp of 5Ј-untranslated region, and 1.7 kbp of 3Ј-untranslated region, was used for further experiments.
As a result of this cDNA screening, we also obtained several Kv␤2 clones. However, none of these clones contained full coding sequences. We again employed reverse transcriptase-PCR using the following primer set (5Ј-primer, B2-5 5Ј-CTGATCTAGATAAGTGAGGC-3Ј; 3Јprimer, B2-3 5Ј-CTATCGATGACTTAGGATCTATAGTCC-3Ј), flanking the entire 1101-bp coding region of Kv␤2, to obtain the coding region of Kv␤2. After 25 rounds of PCR, a specific amplification product of the predicted size was obtained and subcloned into pRBG4 at the XbaI and ClaI sites. Eight cDNAs corresponding to the predicted size were identified by restriction mapping. One of these clones was analyzed by sequencing and was used as Kv␤2/RBG4. We obtained similar results using either the RAT heart (RAK) or brain (BK2) cDNA; the data presented here were obtained with the RAK clone.
Expression and Analysis of K ϩ Channel ␣and ␤-Subunits-Procedures for COS-1 cell culture, DNA transfection, immunoblot analysis, and immunoprecipitation reactions were performed essentially as described in Shi et al. (19) with the following exceptions. For immunoblots, cells were extracted as described previously in 500 l of lysis buffer, and the soluble lysate and insoluble pellet were separated by centrifugation in the microcentrifuge at 15,800 ϫ g for 2 min. The supernatant (lysate) was diluted with an equal volume of 2 ϫ reducing SDS sample buffer. For metabolic labeling in [ 35 S]methionine, cells grown on 60-mm tissue culture dishes were pre-incubated in methionine-free DMEM for 10 min at 37°C followed by incubation in methionine-free DMEM containing 333 Ci/ml of [ 35 S]methionine at 37°C for 2-4 h. Cells were then washed and extracted with 1 ml of lysis buffer. For immunoprecipitation reactions, 100 l of lysate was used, and the resultant products were analyzed on 9% SDS-polyacrylamide gel electrophoresis and visualized by fluorography on Kodak BIOMAX film or by phosphorimaging (Molecular Dynamics).

Expression of ␤-Subunits by Transient Transfection in COS-1 Cells
We have previously characterized the expression of ␤-subunits in rat brain using an anti-␤-subunit antibody (8). This antibody, raised against the conserved C-terminal region of Kv␤1 and Kv␤2 (and Kv␤3), recognizes several polypeptides in rat brain, among these a predominant polypeptide species of 38 kDa, a polypeptide of 41 kDa, and minor species at 44 kDa (Fig. 1, lane 1). As a first step toward correlating these brain polypeptides with the recombinant ␤-subunits, we expressed Kv␤1 and Kv␤2 cDNAs by transient transfection into COS-1 cells (19) and investigated the expressed polypeptides by immunoblots. Surprisingly, a minor immunoreactive polypeptide species of 44 kDa in rat brain membranes comigrates with the 1 The abbreviations used are: DMEM, Dulbecco's modified Eagle's medium; PCR, polymerase chain reaction; bp, base pair(s); kbp, kilobase pair(s).
2 J. S. Trimmer, unpublished data. recombinant Kv␤1 polypeptide (Fig. 1, lane 2). Comigration of the major ␤-subunit immunoreactive polypeptide of 38 kDa with recombinant Kv␤2 (Fig. 1, lane 3) is consistent with our previous proposal (8) that this abundant brain polypeptide is in fact Kv␤2. Similar electrophoretic mobilities for these recombinant ␤-subunits are obtained in two other mammalian cell lines (HEK293, PC12; not shown), suggesting that cell typespecific post-translational modifications do not contribute significantly to the mobility of ␤-subunit polypeptides. This suggests, but does yet not prove, that the prominent 41-kDa immunoreactive band in rat brain is not Kv␤1 and that Kv␤1 apparently corresponds to the 44-kDa polypeptide.

Specific Association of Kv␤1 and Kv␤2 with Kv1.2
To study the selectivity of ␣/␤-subunit interaction, we undertook a biochemical approach utilizing co-immunoprecipitation from cotransfected COS-1 cells. Except where explicitly stated otherwise, all immunoprecipitation reactions were performed under conditions designed to maintain subunit association, resulting in some nonspecific background, even in reactions performed in the absence of antibody. Initially, we focused on ␤-subunit interaction with Kv1.2, based on previous studies in brain (3,8). Fig. 2 shows a fluorographic image of immunoprecipitation products fractionated on an SDS gel. Kv1.2-, Kv␤1-, and Kv␤2-transfected cells express 65-, 44-, and 38-kDa proteins, respectively (Fig. 2), and in each case subunit-specific antibodies show no detectable cross-reactivity to heterologous samples. As expected, Kv␤1 and Kv␤2 are immunoprecipitated with the pan-␤ antibody from Kv1.2/Kv␤1-or Kv1.2/Kv␤2cotransfected cells (Fig. 2). Both ␤-subunits could also be coimmunoprecipitated with the anti-Kv1.2 antibody. The presence of Kv1.2/Kv␤ interaction was confirmed by reciprocal coimmunoprecipitation reactions by the presence of Kv1.2 in the ␤-subunit immunoprecipitation products.
Addition of a denaturing agent, such as the detergents SDS and deoxycholate, should affect polypeptide folding and disrupt the noncovalent protein-protein interactions typical of most multi-subunit membrane protein complexes (20). To test if K ϩ channel subunit association was through similar noncovalent interactions, immunoprecipitation reactions were performed in the presence of such denaturing agents. The coprecipitation of Kv1.2 with Kv␤2 could be disrupted by the addition of 0.2% SDS and 0.5% sodium deoxycholate during the immunoprecipitation reactions; this treatment has no effect on the direct immunoprecipitation of the subunits themselves (Fig. 3A). Similar results were obtained for Kv1.2-Kv␤1 interaction (not shown). Thus, K ϩ channel ␣/␤-subunit interaction has similar sensitivity to denaturing detergents as exhibited for other multisubunit membrane protein complexes (20).
To test whether co-expression within the same cell is necessary for subunit interaction, individually transfected dishes of COS-1 cells expressing either Kv1.2 or Kv␤2 were harvested. The cells were then pooled, and the pooled mixture of cells was extracted under standard conditions. The resultant lysates  3. A, effect of SDS addition to immunoprecipitation reactions. Cells were transfected with Kv1.2 and Kv␤2. Cells were labeled with [ 35 S]methionine for 2 h, harvested in lysis buffer, and the lysates were subjected to immunoprecipitation with anti-Kv1.2 ("␣" lanes) or anti-␤ ("␤" lanes) antibody or without antibody ("Ϫ" lanes). Immunoprecipitation reactions were carried out under the presence (right panel) or absence (left panel) of 0.2% SDS and 0.5% sodium deoxycholate. Numbers on left refer to mobility of prestained molecular weight standards. B, immunoprecipitation from the mixed-cell lysate of individually transfected dishes of cells. COS-1 cells individually transfected with Kv1.2, Kv␤1, or Kv␤2 were mixed and lysed, and the lysates were subjected to immunoprecipitation. The combinations of the singly transfected cells used for immunoprecipitation are indicated at the top of the lanes labeled "mix". Lanes labeled "co" show the results obtained from cotransfected cells expressing the same ␣/␤-subunit combination. The immunoprecipitation reactions were performed with anti-Kv1.2 ("␣" lanes) or anti-␤ ("␤" lanes) or without antibody ("Ϫ" lanes). Numbers on left refer to mobility of prestained molecular weight standards.
were then subjected to immunoprecipitation with subunit-specific antibodies. These experiments yielded no co-immunoprecipitation of ␣and ␤-subunits above background (no antibody lanes), showing that co-expression within the same cell is necessary for subunit interaction (Fig. 3B).
Previous studies had shown that deletion of the N terminus of Kv␤1 destroyed its ability to modulate inactivation (4). To test whether this was simply due to a lack of interaction, an N-terminal truncation mutant, Kv␤1⌬N70, which lacks amino acids 1-70, was co-expressed with Kv1.2. As shown in Fig. 4A, Kv␤1⌬N70 can be efficiently co-immunoprecipitated with anti-Kv1.2 antibody and vice-versa. Thus, removal of the domain necessary for Kv␤1-mediated modulation of inactivation does not disrupt ␣/␤-subunit interaction, showing that the loss of the ability of such mutants to modulate inactivation is not due to an inability to interact with ␣-subunits. A similar N-terminal deletion of Kv␤2 (Kv␤2⌬N22) also exhibited interaction with Kv1.2 that was indistinguishable from wild-type Kv␤2 (Fig.   4B). These data indicate that the N-terminal domains of ␤-subunits are not necessary for the interaction with ␣-subunits and that the interaction domain lies somewhere else in the ␤-subunit sequence.

Selective Association of Kv␤1 and Kv␤2 with ␣-Subunits
To investigate the selectivity of ␣/␤-subunit interaction, coimmunoprecipitation from cells co-expressing pairwise combinations of recombinant mammalian ␣-subunits and Kv␤1 and Kv␤2 was performed. Control experiments, as detailed above, were performed for each set of ␣/␤-subunit combinations. However, due to space limitations, only the relevant co-immunoprecipitation reactions are presented here.
Kv1 Subfamily-Five members of the mammalian Shakerrelated (Kv1) subfamily were tested for interaction with Kv␤1 and Kv␤2. All of the Kv1 family members tested (Kv1.1, Kv1.2, Kv1.3, Kv1.5, and Kv1.6) exhibit direct and specific interaction with both Kv␤1 and Kv␤2, as evidenced by reciprocal co-immunoprecipitation (Fig. 5). However, distinctions are apparent in the extent of co-immunoprecipitation among the specific pairwise combinations. Kv1.3 and Kv1.6 are similar to Kv1.2, in that high levels of co-immunoprecipitation of both the ␣and ␤-subunits are observed in reactions using either anti-␣-subunit or anti-␤-subunit antibody. Kv1.1 and Kv1.5, however, show lower than expected levels of coprecipitated ␣-subunits in the anti-␤-subunit immunoprecipitation reactions, perhaps due to an overabundance of ␤-subunits such that only a small fraction of the large total ␤-subunit pool is associated with the small ␣-subunit pool. However, it is not possible to determine from these types of experiments whether these differences reflect quantitative differences in ␣/␤-subunit association. Control experiments on singly transfected cells expressing ␣-subunits alone show no detectable immunoprecipitation with anti-␤-subunit antibodies. This verifies that the low levels of Kv1.1 and Kv1.5 seen in anti-␤-subunit immunoprecipitation reactions performed on cells co-expressing ␣and ␤-subunits are specific and significant.
Kv2 and Kv3 Subfamily-Analysis of cells cotransfected with Shab-related Kv2.1 and either Kv␤1 or Kv␤2 show no coprecipitation by anti-␤ subunit antibody (Fig. 6A). Low levels of Kv␤1 and Kv␤2 are seen in immunoprecipitation reactions performed with the anti-␣-subunit antibody; however, comparable levels are observed in similar immunoprecipitation reactions performed on cells expressing ␤-subunits alone (not shown), indicating that these products are due to minor crossreactivity of the anti-Kv2.1 antibody to these ␤-subunits and not to ␣/␤-subunit interaction. Similar nonspecific immunoprecipitation of low levels of ␤-subunits was also seen in immunoprecipitation reactions performed on cotransfected cells in the absence of antibody (not shown). The addition of denaturing agents, SDS and deoxycholate, during the immunoprecipitation reactions shows that this treatment has no or very weak effect on the relatively low but detectable level of co-immunoprecipitation (Fig. 6B). Taken together, these data indicate that the observed co-immunoprecipitation is not due to specific noncovalent interactions between Kv2.1 and Kv␤2, as these sorts of intermolecular associations are typically disrupted by denaturing agents (see Fig. 3A) but is due to low levels of antibody cross-reactivity or other nonspecific precipitation. When these ␤-subunits were co-expressed with the mammalian Shaw homolog Kv3.1, no detectable co-immunoprecipitation was observed (Fig. 6C), although strong subunit-specific immunoprecipitation was observed.
Kv4 Subfamily-Strong reciprocal co-immunoprecipitation was observed between both Kv␤1 and Kv␤2 and the mammalian Shal homolog Kv4.2 (Fig. 7A). The interaction of Kv␤2 and Kv4.2 is relatively resistant to treatment with the denaturing detergent SDS (SDS treatment) in that the co-immunoprecipitation is not disrupted by the addition of SDS at concentrations less than 0.6% (Fig. 7B). This is distinct from the characteristics of the interaction of Kv1.2 with Kv␤1 and Kv␤2, where interaction is partially disrupted by the addition of SDS to only 0.2%, with complete disruption observed at 0.4% SDS (Fig. 7B).

DISCUSSION
Our previous study using an antibody against a sequence conserved in both Kv␤1 and Kv␤2 revealed the existence of multiple immunoreactive ␤-subunits in rat brain (8). Here, analysis of transfected cells expressing recombinant Kv␤2 and Kv␤1 reveals that a minor 44-kDa rat brain ␤-subunit comigrates with Kv␤1, while the major 38-kDa ␤-subunit comigrates with Kv␤2. The other immunoreactive ␤-subunit at 41 kDa, which is recognized by the ␤-subunit antibody, is apparently neither Kv␤1 nor Kv␤2 and suggests the existence of an additional, as yet uncharacterized member of the ␤-subunit gene family in rat brain. Recent cloning of a partial cDNA for a rat Kv␤3 ␤-subunit, which shares the same nucleotide sequence with Kv␤1 except for its unique N-terminal region and is predicted to encode a polypeptide of 45 kDa, strongly suggests the presence of at least one alternatively spliced product of the Kv␤1 gene (5). Studies with subtype-specific antibodies will allow for the eventual unequivocal identification and localization of each of the individual components of the ␤-subunit pool in brain. Moreover, comprehensive molecular analysis of the ␤-subunit gene family will lead to the identification of other ␤-subunits, for instance, those associated with Shab-(see Ref. 12), Shaw-, and Shal-(21) related K ϩ channels.
We found that Kv␤1 and Kv␤2 expressed in COS-1 cells could associate with all five of the Shaker-related subfamily members tested (Kv1.1, Kv1.2, Kv1.3, Kv1.5, Kv1.6), as well as with the Shal-related Kv4.2. The ratio of the amount of coimmunoprecipitation seen on anti-␣ and anti-␤ lanes exhibited some variation among the different ␣/␤ pairwise combinations. This discrepancy in the extent of reciprocity of co-immunoprecipitation could be due to the relatively low efficiency of immunoprecipitation with anti-␤ antibody due to the high expression level of ␤-subunit in the cotransfected cells. Kv2.1 and Kv3.1 exhibited no detectable co-immunoprecipitation, suggesting these two ␣-subunits are unable to interact with Kv␤1 and Kv␤2. Subcellular localization of Kv2.1 and ␤-subunits in transfected cells is consistent with this model in that immunofluorescence staining of cells co-expressing Kv2.1 and Kv␤2 shows no overlap of ␣and ␤-subunit staining, while cells co-expressing Kv1.2 and Kv␤2 with Kv␤2 show extensive overlap throughout the cells. 3 Our results provide direct biochemical evidence for selective interaction of K ϩ channel ␤-subunits with only a subset of the ␣-subunit gene family, have greatly expanded the initial observations of Rettig et al. (4) who showed that Kv1.1 and Kv1.4 interact functionally with Kv␤1 in oocytes (4), and provide the evidence for a direct, noncovalent interaction between ␣and ␤-subunits. These results also confirm and extend our previous studies of rat brain ␣/␤-subunit association in situ, where we found that neuronal ␤-subunits could be coprecipitated with rat brain Kv1.2 and Kv1.4 but not with Kv2.1 (8). A detailed characterization of purified bovine brain dendrotoxin acceptors, which were later found to contain Kv␤1 and Kv␤2 (3), showed that these K ϩ channel complexes contain Kv1.1, Kv1.2, Kv1.4, and Kv1.6 (22). Our findings provide a first step toward understanding the molecular determinants of ␣/␤-subunit interaction by showing that the subunit selectivity observed in rat brain can be recapitulated in transfected cell lines, indicating that selectivity is mainly determined by the primary structure of the interacting subunits.
The voltage-gated K ϩ channel ␣-subunit genes segregate into four subfamilies based on the similarity of primary structure of each member (23). As discussed above, our results show that Kv␤1 and Kv␤2 interaction seemed to be restricted to Shakerand Shal-related subfamilies. Interestingly, proposed phylogenetic trees place the Shaker (Kv1) and Shal (Kv4) subfamilies on one major branch, while Shab (Kv2) and Shaw (Kv3) members are placed on a separate branch (1,24). Thus, the ability to interact with Kv␤1 and Kv␤2 appears to reside in the relatedness of their primary sequences as evidenced by their phylogenetic grouping and allows for the design of structure-function analyses aimed at defining the domains of ␣-subunits mediating ␣/␤-subunit interaction. In the case of voltage-sensitive Ca 2ϩ channel ␣/␤-subunit interaction, the ␤-subunit binds to a conserved cytoplasmic motif in the ␣ 1 -subunit (25). Taken together with the fact that K ϩ channel ␤-subunits are also cytoplasmic proteins, it is likely that the interaction domain on K ϩ channel ␣-subunits is present on a cytoplasmic domain.
No distinct domain of ␣-subunits stands out as a clear candidate for mediating interaction with ␤-subunits based simply on the positive interaction of both Kv1 and Kv4 family members. However, our experiments using SDS treatment to disrupt ␣/␤-subunit interaction imply that the interaction of ␤-subunits with Kv1.2 and Kv4.2 are somewhat distinct. In addition, only Kv1 and not Kv2, Kv3, or Kv4 subfamily ␣-subunits have been found associated with Kv␤1 and Kv␤2 in rat brain in situ (8). 4 Together, these data may imply that the only physiologically relevant subunit interactions are between Kv1 (Shaker-related) ␣-subunits and Kv␤1 and Kv␤2. Using this assumption, a conserved N-terminal, presumably cytoplasmic domain of about 130 amino acids is striking in that it is highly conserved among Kv1 ␣-subunits but not among members of the other (Kv2, Kv3, and Kv4) subfamily members. This highly conserved region, known as the "T1" (26) or "NAB" (27) domain, is thought to be important in mediating efficient ␣/␣-subunit interaction (26 -28). This may raise the interesting scenario whereby both ␣/␣and ␣/␤-subunit interactions are mediated through similar domains. Extensive mutational analysis of ␣-subunit proteins will lead to the elucidation of the specific ␤-subunit binding region.