KChIP3 rescues the functional expression of Shal channel tetramerization mutants.

KChIP proteins regulate Shal, Kv4.x, channel expression by binding to a conserved sequence at the N terminus of the subunit. The binding of KChIP facilitates a redistribution of Kv4 protein to the cell surface, producing a large increase in current along with significant changes in channel gating kinetics. Recently we have shown that mutants of Kv4.2 lacking the ability to bind an intersubunit Zn(2+) between their T1 domains fail to form functional channels because they are unable to assemble to tetramers and remain trapped in the endoplasmic reticulum. Here we find that KChIPs are capable of rescuing the function of Zn(2+) site mutants by driving the mutant subunits to assemble to tetramers. Thus, in addition to known trafficking effects, KChIPs play a direct role in subunit assembly by binding to monomeric subunits within the endoplasmic reticulum and promoting tetrameric channel assembly. Zn(2+)-less Kv4.2 channels expressed with KChIP3 demonstrate several distinct kinetic changes in channel gating, including a reduced time to peak and faster entry into the inactivated state as well as extending the time to recover from inactivation by 3-4 fold.

The formation of voltage-gated potassium (Kv) 1 channels is a multistep process with many different interactions and folding events required to form the completed channel (1). The common functional core of all Kv channels assembles as a tetramer of pore-forming ␣-subunits. This tetramer is the core of the future ion channel signal transduction complex, but additional folding steps as well as interactions with auxiliary proteins occur before the final functional channel complex at the cell surface is formed. Many auxiliary subunit proteins that bind to Kv ␣-subunits have been identified, but precisely when these interactions occur during channel complex formation and what role these interactions play in helping the channels to assemble, traffic, and function are topics of great interest (1)(2)(3)(4). Through the use of heterologous expression systems and mutagenesis studies, we can expose many of these important interactions and folding events, and reveal the processes by which Kv channel complexes form. A comparison of channel expression and functional properties with and without specific auxiliary proteins reveals how these different processes contribute to the formation and function of ion channel complexes.
An early step in Kv channel formation involves the tetramerization of the ␣-subunit T1 domains at the cytoplasmic N terminus of the protein (5-7). For Kv4.2 channels, a critical component of the T1 domain interaction involves the coordination of an intersubunit Zn 2ϩ ion found on non-Shaker type Kv channel T1 domains (8 -10). Although Zn 2ϩ binding sites are common in proteins, intersubunit Zn 2ϩ binding sites, as found in the T1 domain, are relatively rare. To determine what functions might be regulated by the T1 intersubunit Zn 2ϩ site, we generated a series of mutations to the Zn 2ϩ coordination residues and tested them for cell surface expression (8). We found that mutation to any of the Zn 2ϩ binding site amino acids caused a block of functional channel expression by disrupting subunit tetramerization. This disruption traps the protein within the endoplasmic reticulum and produces a shift in the migration of solubilized subunits on size exclusion chromatography to monomeric molecular weights (8). Therefore, we can clearly say that one important function of the Zn 2ϩ site is to stabilize channel assembly. Because functional channels could not be formed without the intact Zn 2ϩ site, we sought to determine if it was possible to rescue these mutant channels by co-expression with KChIP auxiliary subunit proteins that are known to enhance Shal channel expression by 10 -50 times (3).
Kv4.2 subunits form channels in the absence of KChIP auxiliary proteins; however, co-expression with KChIP enhances current expression by altering the trafficking of Kv4 channels (11). KChIP auxiliary subunits interact with Kv4.2 channels by binding to the first 14 residues of the cytoplasmic N terminus and possibly through direct binding interactions with the T1 domain (3,(12)(13)(14) containing the zinc site. Also, the KChIP subunit itself is capable of multimeric interactions following binding to the Kv4 N terminus, providing an alternative and non-zinc-dependent pathway to channel complex multimerization (12). The KChIP enhancement of expression has been attributed to a trafficking action of KChIPs, causing channel proteins to redistribute to the cell surface (3,11). The mechanism responsible for this redistribution is not known. KChIPs could be acting to enhance channel assembly, enhance folding of the assembled channels, block the action of a retention sequence on Kv4.2 N terminus, providing a stronger surface expression motif to the channel complex, or a combination of effects (1). In addition to increasing expression, KChIP binding to the Kv4.2 N terminus modulates the functional gating properties of the channels. The most dramatic effects are on the inactivation properties of the channels, where KChIP drastically slows inactivation and speeds the recovery from inactivation. Inactivation is an important feedback mechanism that limits Shal channel function at depolarized potentials and gives these channels their characteristic "A" current waveform. The N terminus of Kv4.2 is clearly involved in inactivation gating, since deletion of this region produces slowed inactivation. Moreover, experiments on the functional properties of the N terminus show that it can act as a pore-binding peptide (15)(16)(17). Therefore, the gating changes produced by KChIPs might depend on the binding and sequestering of the Kv4 N terminus, causing an alteration to a complex set of allosteric interactions that involve other structural elements of the channel (18). Here we have tested whether KChIP3 binding promotes assembly of the Zn 2ϩ site mutant Kv4.2 subunit proteins by testing for the rescue of channel tetramerization and cell surface function by KChIP3. We then characterized the resultant complexes as to their assembly and biophysical properties in comparison to wild-type Kv4.2 co-expressed with KChIP3.

EXPERIMENTAL PROCEDURES
DNA Subcloning-Rat Kv4.2 cDNA was obtained as a gift from the L. Jan laboratory (pBS-rKv4.2). The Kv4.2 coding region was transferred to a cytomegalovirus-based vector (Clontech, Inc.), and also used to generate a vector for bacterial expression of the T1 domain, as described previously (8). KChIP3 was obtained as an EST clone, 2403205, and inserted into the same vector using available XhoI ϩ NotI restriction sites.
In addition, as previously described (8), zinc coordinate site mutants were made using the QuikChange™ strategy and reagents (Stratagene, Inc.), with oligonucleotides that were 36 -48 bases in length. All mutations were confirmed by DNA sequence analysis through the entire coding region.
Cell Culture-CHO-K1 cells were obtained from ATCC and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 40 g/ml L-proline, 100 units/ml penicillin, and 100 units/ml streptomycin.
Transfection-CHO cells were transiently transfected with cDNAs using Lipofectamine, as recommended by the manufacturer (Invitrogen Life Technologies, Inc.). Enhanced GFP (EGFP, Clontech) was coexpressed with the channel to assess transfection efficiency and to identify expressing cells for voltage clamp experiments (26). One day prior to transfection, cells were plated onto cover slips. At 18 -24 h, confluency of the cells was about 50 -60% and considered optimal for transfection. We used a mixture of 0.5 g of cDNA (Kv4.2 or zinc site mutants), 0.5 g of GFP, and 3 l of Lipofectamine in 0.5 ml of serumfree Dulbecco's modified Eagle's medium for about 2 h after which regular medium was restored. All the recordings were done 18 -20 h after transfection.
For the Kv4.2 and KChIP3 ratio study, total cDNA used per transfection was fixed at 4.5 g. The Kv4.2 cDNA level was fixed at 0.8 g and EGFP cDNA at 0.5 g in the transfection mix. Different amounts of KChIP3 were used to generate the different expression ratios. Red fluorescent protein cDNA was used as a filler to keep the cDNA total amount fixed. Transfection was carried out as described above. Recordings were done at 18 -20 h after transfection. For the comparative studies on the effects of different KChIP3 ratios, the experiment was performed identically with the zinc triple mutant (Kv4.2(ϪZnB3)) cDNA level also fixed to 0.8 g. Because of the limited cDNA capacity of our transfection reagent, this experimental design limited us to a maximal Kv4.2 to KChIP3 cDNA ratio of 4:1. Therefore we could not test whether the mutant expression had completely plateaued. However, a Student's t test between wild type and mutant at this cDNA ratio revealed no significant difference in the expression level.
Whole cell currents were sampled at 5 kHz and filtered at 2 kHz (Ϫ3 dB, four pole Bessel filter). Series resistance was compensated by 90 -95%. Data were leak-subtracted online using P/8 protocol. The holding potential was Ϫ80 mV. Records were digitized with a Digidata-1200B A/D converter (Axon Instruments) and stored directly on a hard disk using pClamp version 6.0.5. All recordings were done at room temperature (25 Ϯ 2°C).
Size Exclusion Chromatography-Transfected cells were solubilized in CHAPS (2% final concentration) and run on a Superose 6 column using an Amersham Biosciences FPLC system as described previously (8). Fractions were run on SDS-PAGE gels and Western blotted to determine the fractions containing the proteins of interest. Molecular weight standards were run and the standard curve corrected to the appropriate fraction by the dead time between the detector and collector. Kv4.2 protein often runs as a doublet on SDS-PAGE gel, with a fraction of the protein running at a high molecular size. There was no difference in the FPLC profiles for these two bands, and the higher molecular size band appeared to be protein that either remained tetrameric or aggregated in the running of the gel. The Western blots in Fig. 3 only show the profile for the immunoreactive material at the correct size for Kv4.2, ϳ70 kDa.
Data Analysis-Data analysis was performed using Clampfit version 6.0.5 (Axon Instruments) and Origin 6.1 software (Microcal Software, Inc.). Pooled data are expressed as mean ϩ S.E. Statistical comparisons between groups of data were carried out with the two-tailed Student's t test for paired or unpaired data or 1 way ANOVA using Origin and Analyze-it for Excel (Analyze-it Software). Tests of significance were performed based on standard hypothesis testing approaches. The analysis method of choice depended upon the null hypothesis to be tested: comparison of two variables or sets of variables. Student's t tests were performed to determine if a single measured variable was significantly different between mutant and wild type. ANOVA comparisons were performed to test if a variable range was significantly different between mutant and wild type; for example, if mutants produce overall changes in the time constants of inactivation. Similarly ANOVA comparisons were used to test whether a channel property, such as inactivation time constant, showed significant changes with voltage for mutant or wild type. Values of p Ͻ 0.05 were considered statistically significant.
Reversal Potential Measurement-Cells were stepped to ϩ50 mV for 15 ms and then to different potentials from Ϫ140 mV to ϩ50 mV at 10 mV steps. The holding potential was at Ϫ80 mV. An I-V curve was plotted with the peak tail current as y-axis and command potentials as x-axis. The potential where the peak tail current intersected the x-axis was recorded as the reversal potential. Under typical recording conditions, reversal potential was measured as approximately Ϫ90 mV.

KChIP3 Rescues Function of Zn 2ϩ Site Mutant-
The interaction of KChIP auxiliary subunits with Kv4.2 ␣-subunits is an important step in the formation of functional A-type Kv channels (3,11). We were interested in testing if the intersubunit interactions of KChIPs might also have a direct role in channel assembly beyond their established role in redistributing channels onto the cell surface. In order to test this hypothesis, we sought to determine if KChIP co-expression could rescue the loss of A-current produced by Zn 2ϩ site assembly mutants of Kv4.2. For rat Kv4.2, the residues involved in coordination of Zn 2ϩ are His 105 , Cys 111 , Cys 132 , and Cys 133 (8). Our previous studies have shown that mutations to these sites eliminate functional channel expression by disruption of subunit assembly and trapping of monomeric subunits on intracellular membranes (8). For these studies, we have adopted the following nomenclature to refer to these mutations: single mutants are indicated by the mutation in parentheses after the construct name; multiple mutants are called Kv4.2(ϪZnB) for "minus Zn 2ϩ binding," with a specific number to indicate the sites mutated. Thus, Kv4.2(ϪZnB2) has the mutations C132A and C133A. Kv4.2(ϪZnB3) has the additional mutation C111A, and Kv4.2(ϪZnB4) has the additional mutation H105A. In general, we have focused our studies on the Kv4.2(ϪZnB3) mutant, lacking all 3 coordinating Cys residues, but have performed complementary expression, assembly, and biophysical studies on other ϪZnB mutants and EGFP-tagged versions of these channels, and found that they behave similarly.
As expected from our previous results, wild-type Kv4.2 expression produces A-currents in CHO cells (Fig. 1A); however, the Kv4.2(ϪZnB3) mutant showed no functional expression without KChIP co-expression ( Fig. 1B and Ref. 8). With co-transfection of KChIP3 at a 1:1 cDNA ratio, wild-type Kv4.2 shows a large increase in functional A-current (see Fig.  1C). Interestingly, co-expression with KChIP3 rescues Kv4.2(ϪZnB3) mutant subunits, producing functional A-current; however, the level of current seen is only ϳ25% of that seen with wild-type Kv4.2 channels (Fig. 1D). We confirmed this basic finding with several other Zn 2ϩ site mutants, and the summary results are presented in Fig. 2A. Although none of the Zn 2ϩ mutants functioned without KChIP3, they were all rescued to a similar degree, ϳ25% of wild type, when co-expressed with KChIP3 at a 1:1 cDNA ratio.
To determine whether this reduced expression level could be further enhanced with higher levels of KChIP3 expression, we boosted the KChIP3 cDNA levels to be 4ϫ greater than Kv4.2 and retested for A-current levels. Under these conditions, there was no change in wild-type Kv4.2 expression (Fig. 1E); however, the Kv4.2(ϪZnB3) current is now 4ϫ larger, reaching the same level of functional expression seen with wild-type channels (Fig. 1F).
Higher KChIP3 Levels Are Required to See Zn 2ϩ Mutant Expression-The previous results suggest that there is a distinct dose response relationship for KChIP3 enhancement of Kv4.2 channels versus Kv4.2(ϪZnB3). To further explore this effect, we kept the channel cDNA level constant and varied the amount of KChIP3 cDNA, keeping the total amount of cDNA constant. Fig. 2B shows the results of a dose response experiment using different ratios of wild-type Kv4.2 and KChIP3 cDNA. With wild-type Kv4.2, significant increases in current are seen with a 1:30 ratio of KChIP3 to Kv4.2 ( Fig. 2B; ratio ϭ 0.033). At a 1:10 ratio ( Fig. 2B; ratio ϭ 0.1), the enhancement of expression is maximal, and no further increases in current are seen with higher levels of KChIP3 cDNA. The 10-fold lower requirement for KChIP3 cDNA is likely caused by more efficient synthesis of this soluble protein compared with the synthesis of the polytransmembrane channel protein.
With Kv4.2(ϪZnB3), the dose response relationship is significantly different, with additional KChIP cDNA required for maximal current expression. No expression of channels was observed with a 1:10 ratio of KChIP3 to Kv4.2(ϪZnB3) (see Fig.  2B; ratio ϭ 0.1). Expression was first evident at a ratio of 1:4 ( Fig. 2B; ratio ϭ 0.25), with increasing current at increasing KChIP3 to Kv4.2(ϪZnB3) ratios until functional expression levels comparable to wild-type channels were achieved at a KChIP3 to Kv4.2 ratio of 4:1 ( Fig. 2B; ratio ϭ 4.0). Although, the cDNA ratios obviously do not accurately reflect the amount of protein being expressed, the 40-fold right shift in the curve with Kv4.2(ϪZnB3) suggests that a mass action effect is occurring, where KChIP3 binding to Zn 2ϩ site mutant channel subunits is driving functional expression of these channels. These results strongly support our previous hypothesis that the loss of function produced by Zn 2ϩ site mutations is a specific problem in the assembly of the channel rather than a complete nonspecific disruption of the subunit protein, and suggest that the primary action of KChIP3 is to overcome this assembly defect (8,10). To further confirm this conclusion, we sought to determine whether the assembly state of the channel had indeed changed following KChIP3 co-expression.
Tetramerization of Zn 2ϩ Site Mutants-Our previous studies have shown that the Kv4.2(ϪZnB) subunit protein runs as a monomer on SEC-FPLC compared with the normal tetrameric position of the wild-type channel (8). In order to test the hypothesis that KChIP3 was rescuing the function of Kv4.2(ϪZnB) by driving tetramer assembly, cells were transfected with KChIP3 and Kv4.2, either with or without Zn 2ϩ site mutations, and then the proteins were solubilized and the channel assembly state characterized by SEC. In the absence of KChIPs, wild-type Kv4.2 runs in fractions consistent with its self-assembly into tetramers, whereas Kv4.2(ϪZnB3) is found in lower molecular size fractions consistent with monomers (see Fig. 3A and Ref. 8).
Following co-expression of wild-type Kv4.2 with KChIP3, the channel remains in the tetrameric fractions (Fig. 3B). KChIP3 is found to co-migrate with the channel in addition to the normal low molecular mass position seen for unassembled KChIP3, indicating a stable incorporation into the channel complex (see Fig. 3C). Although KChIP addition is expected to increase the channel protein mass by ϳ25%, there was no large change in the position of the channel protein in the fractionation, suggesting that KChIP3 is tightly integrating into the structure and not significantly increasing the hydrodynamic radius. Support for this conclusion comes from single particle reconstruction studies on Kv4.2 channels co-expressed with KChIP2, where the KChIP2 density is found to be tightly integrated with the T1 domain and the cytoplasmic C terminus (19).
For Kv4.2(ϪZnB3), the results of forming a channel complex with KChIP3 are much more dramatic. Expressed alone, Kv4.2(ϪZnB3) is found exclusively in the monomeric fractions (see Fig. 3A). Co-expression with KChIP3 at a 1:1 cDNA ratio (Fig. 3B) produces a partial redistribution of Kv4.2(ϪZnB3), where both tetrameric and monomeric channel protein is seen on SEC-FPLC. If the KChIP cDNA level is further boosted to 4ϫ the level of Kv4.2(ϪZnB3) cDNA, then the shift to tetramer is complete with all protein found in the tetrameric fractions (4ϫ in Fig.  3B). These results on assembly state correlate well with the expression level dependence for KChIP3 rescue of Kv4.2(ϪZnB3) function. Co-expressed KChIP3 is stably incorporated into the  ϪZnB3) and KChIP3 cDNA rescues mutant function, resulting in significant levels of A-current, ϳ25% less than seen with wild-type channels. E, elevation of the level of KChIP3 cDNA to 4ϫ higher than Kv4.2 cDNA does not further enhance current levels. F, elevation of KChIP3 cDNA to 4ϫ higher than Kv4.2(ϪZnB3) cDNA results in complete rescue of current to wild-type levels.
channel complex since it co-migrates with this tetrameric Kv4.2(ϪZnB3); however, the majority of the KChIP3 protein is found in the free KChIP fractions, as expected for our mass action hypothesis (Fig. 3C). Similar results were obtained with EGFPtagged versions of the ZnB3 mutant, (ϩEGFP)Kv4.2(ϪZnB3) (Fig. 3, A-C). We also confirmed with several other Zn 2ϩ site mutant constructs that co-expression of KChIP3 with mutant channels at a 4:1 ratio produces shifts in channel protein migration from monomer to tetramer fractions (data not shown). These results strongly support our hypothesis that the action of KChIP3 in the rescue of Kv4.2(ϪZnB) mutants overcomes an assembly defect and drives the subunits to tetramer even in the absence of the Zn 2ϩ binding interactions that would normally stabilize the T1 tetrameric interaction.
Trafficking of Zn 2ϩ Mutants out of the Endoplasmic Reticulum by KChIP3-Our results have shown that KChIP3 coexpression rescues the function of Zn 2ϩ binding site mutants by driving the assembly of these subunits to tetramer. We next confirmed that in the presence of KChIP3 the Zn 2ϩ site mutant channel protein was released from the endoplasmic reticulum to allow trafficking to the cell surface. For these studies we have used COS-7. COS-7 cells are large flat cells that provide a beautiful view of the internal membrane systems of the cell. In Fig. 4A, we show the distribution of EGFP-tagged Kv4.2 protein, (ϩEGFP)Kv4.2, in COS-7 cells in the absence of KChIP3. Without KChIP3, the protein is strongly retained in endoplasmic reticulum and Golgi membranes, as can be seen by strong fluorescence within the internal membrane systems of the cell (8,11). Addition of low or high levels of KChIP3 redistributes wild-type Kv4.2 protein out of these internal membrane producing a diffuse fluorescence over the entire cell, as expected for proteins reaching the surface membranes of these flat cells (Fig. 4C). In Fig. 4B, we show a cell that expresses (ϩEGFP)Kv4.2(ϪZnB3) subunits alone. As described previously, similar to wild-type channels, the Zn 2ϩ site mutant channels are retained intracellularly (8). For Kv4.2(ϪZnB3), addition of low levels of KChIP3 cDNA to the transfection does not noticeably change the protein distribution; however, with a 4ϫ higher level of KChIP3 cDNA, the protein is clearly released from the internal membrane and produces a diffuse fluorescence pattern, similar to wild-type channels (Fig. 4D).
Functional Properties of Zn 2ϩ -less Kv4 Channels-The ability to rescue Kv4.2(ϪZnB) function by co-expression with KChIP3 provides us with an opportunity to characterize the changes in channel functional properties without Zn 2ϩ bound to the Kv4.2 T1 domain. Because we were unable to express Kv4.2(ϪZnB) without KChIP co-expression we focused our studies on answering the question of how the A-channels produced by co-expression of KChIP3 and Kv4.2(ϪZnB) constructs are different from channels produced by co-expression of KChIP3 and Kv4.2.
In Fig. 5A, we have plotted voltage-clamped current traces for Kv4.2 and Kv4.2(ϪZnB3) co-expressed with KChIP3 in response to voltage steps from Ϫ30 mV to ϩ50 mV. The currents are normalized to the same peak current, and the traces overlaid to allow a general comparison of the waveforms for normal and mutant channels. The first effect that is quite obvious is that the A-currents decay more rapidly for the Kv4.2(ϪZnB3) mutant channels than wild type. We have examined other Zn 2ϩ site mutants and find this to be a general property of these channels: loss of Zn 2ϩ binding accelerates the inactivation of channels during sustained depolarizations. A second effect is seen by examination of the currents in response to the smallest shown depolarization to Ϫ30 mV, which shows a higher level of current for wild-type channels than for Kv4.2(ϪZnB3) channels (Fig. 5A). Finally, examination of the traces at higher sweep speeds shows that the Kv4.2(ϪZnB3) currents rise, reach their peak, and begin decaying before wildtype Kv4.2 currents (see Fig. 5A, inset).
We next characterized the steady state inactivation and peak activation properties for normal and Zn 2ϩ site mutant channels. Steady state inactivation curves were measured for both FIG. 2. KChIP3 rescues Zn 2؉ site mutants. A, functional peak current expression levels with 1:1 co-expression of KChIP3 cDNA with different Zn 2ϩ site mutants of Kv4.2 in CHO-K1 cells recorded at ϩ50 mV. Currents normalized by cell size as measured by cell capacitance. Summary data shown for all constructs with and without KChIP3 addition. Only wild-type Kv4.2 shows any functional current without KChIP. All 5 tested Zn 2ϩ site mutants were functionally rescued to a level of expression ϳ25% of that seen with wild-type Kv4.2 co-expressed with KChIP3. A similar rescue of function is seen with EGFP-tagged versions of these mutants (data not shown). One-tailed Student's t tests compared with wild type showed that expression of all mutants were significantly reduced compared with wild type. A one-way ANOVA between the mutants revealed no significant difference between expression levels of different mutants. B, dose response for KChIP3 increase in current expression is right-shifted with Kv4.2(ϪZnB3) mutants. CHO-K1 cells were expressed with a constant amount of channel cDNA and differing ratios of KChIP3 cDNA. The total level of cDNA was kept constant in all experiments. Currents were recorded 24 h after transfection. Wild-type Kv4.2 currents (open squares) peak with KChIP3 cDNA at one-tenth (ratio ϭ 0.1) the level of channel cDNA. Mutant Kv4.2(ϪZnB3) currents (filled squares) do not reach the same level of expression until the KChIP3 cDNA is 4ϫ greater than channel cDNA(ratio ϭ 4.0). Boltzmann curves were fitted to the data sets and the half-maximal response cDNA ratio, R1 ⁄2 , was measured for both Kv4.2 and Kv4.2(ϪZnB3). The measured R1 ⁄2 for KChIP3 expression of the Kv4.2(ϪZnB3) mutant is 40ϫ greater than for wild-type Kv4.2. One-way ANOVA revealed significant differences between the curves for mutant and wild type. A Student's t test performed on the measured half-maximal response cDNA ratio showed a significant difference between mutant and wild type. normal and Zn 2ϩ mutant channels in the presence of KChIP3, see Fig. 5B. For wild-type Kv4.2, the midpoint for the inactivation curve was Ϫ51.1 Ϯ 1.2 mV in the presence of KChIP3. For Kv4.2(ϪZnB3) half-inactivation occurred at Ϫ58.6 Ϯ 0.7 mV indicating an enhanced tendency to inactivate for these mutant channels. Peak activation curves were measured for both normal and Zn 2ϩ mutant Kv4.2 channels expressed in the presence of KChIP3 (Fig. 5B). The activation midpoint of Kv4.2 wild-type channels is Ϫ27.8 Ϯ 0.4 mV compared with Ϫ25.1 Ϯ 0.2 mV for the mutant Kv4.2(ϪZnB3). The activation of Kv4.2(ϪZnB3) channels also showed a slightly steeper slope of 6.0 Ϯ 0.2 compared with 7.0 Ϯ 0.3 for wild-type Kv4.2 channels expressed with KChIP3. Whereas these effects on activation are small there is a significant difference in the overlap between activation and inactivation curves for mutants versus wild type. The decreased overlap for Kv4.2(ϪZnB) channels is expected to produce a much smaller "window" of sustained current between Ϫ60 mV and Ϫ30 mV.
Channel Kinetic Properties-To quantify the kinetic differences in the current waveforms that underlie these differences in function between Kv4.2 channels and Kv4.2(ϪZnB3) chan-nels expressed with KChIP3, we measured the time to peak activation and characterized the inactivation kinetics. For the time required to peak (see Fig. 6A), the Kv4.2(ϪZnB3) channel was consistently faster than wild type, particularly at more depolarized potentials. At ϩ50 mV, the time to peak for Kv4.2(ϪZnB3) channels was 4.3 Ϯ 0.3 ms compared with 7.7 Ϯ 0.8 ms for wild-type channels.
We next examined the changes in inactivation kinetics that produce the more rapid current decay in Zn 2ϩ site mutant channels. Kv4 channel inactivation has been dissected in numerous biophysical studies into three separate processes: open state, closed state, and slow inactivation (15,20,21). Open state inactivation utilizes a pore block produced by the Kv4 cytoplasmic N terminus and is lost by deletion of the Kv4 N terminus or binding of KChIP proteins to the N terminus (16,22). Closed state inactivation is a separate process from open state inactivation. In contrast to open state inactivation, closed state inactivation is not lost by removal of the Kv4 N-terminal inactivation peptide (21). Although the precise mechanism responsible for this inactivation is not known, it is sensitive to mutations within the channel vestibule and results in an apparent uncoupling between S4 movements and channel pore opening (20 -23). A third slower inactivation process also occurs with Kv4 channels that is evident with longer depolarizations and is thought to be a distinct process from C-type inactivation seen in Shaker channels; however, the mechanism for this slow inactivation is poorly understood (21). In the presence of KChIPs, the decay of the current for both normal and Zn 2ϩ site mutants can be as well fit as the sum of 2 exponentials (see Fig. 6B). The faster exponential process is proposed to be dominated by the entry of flickering Kv4.2 channels into closed state inactivation, and the slower process is dominated by movement into the slow inactivated state (16,22). From these fits, we see that loss of the Zn 2ϩ site slightly accelerates the kinetics of both inactivation components as well as shifting the relative weight of the two components toward the faster kinetic (Fig. 6, C and D). These effects all combine to accelerate the decay of A-current for the mutant channels.
Low Voltage Inactivation-To explain the shifts in steady state inactivation produced by Zn 2ϩ site mutations, we need to examine the time course for entry into the low voltage inactivated state as well as the time course for the recovery from inactivation. For Kv4 channels, the inactivation at negative potentials, near the steady state V1 ⁄2 , is dominated by entry into the closed state inactivation process, because it occurs at potentials where channel opening rarely occurs (20,21,23). In Fig. 7, we examined the kinetics for inactivation for wild type as well as Zn 2ϩ mutant channels expressed with KChIP3 at potentials near the V1 ⁄2 for steady state inactivation. In Fig. 7A, we show the closed state inactivation of wild-type Kv4.2 expressed with KChIP3, with an interpulse potential of Ϫ55 mV. Fig. 7B, shows the closed state inactivation of Kv4.2(ϪZnB3) channels with KChIP3 in response to the same inactivation protocol. As expected from the shifts in steady state inactivation, the Kv4.2(ϪZnB3) channels inactivate more completely than wild-type Kv4.2 channels when stepped to the same potential. Average responses are compared in Fig. 7C for inactivation to the two potentials that bracket the V1 ⁄2 for inactivation of these channels. Two differences between wild-type and Kv4.2(ϪZnB3) channels is evident. The first is that Zn 2ϩ mutant channels reach the same level of inactivation at more negative potentials than wild type, as expected from the shift in steady state inactivation. The second is that the kinetics for entry into the closed inactivated state are different between wild-type and Zn 2ϩ mutant channels, inactivating to the same level. At potentials near the V1 ⁄2 , Zn 2ϩ site, mutant channels . Lower number fractions, eluting sooner from the column, contain proteins with a larger hydrodynamic radius. For different conditions, the set of fractions that are expected to contain tetrameric channels are boxed in black, fractions that are expected to contain monomeric subunits are boxed in gray, and fractions that are expected to contain free KChIP3 are boxed in dashes. A, channels expressed without KChIP Western-blotted to identify fractions containing Kv4.2 subunits. Wild-type Kv4.2 channels run in the tetrameric fractions even without KChIP3; however, Kv4.2(ϪZnB3) channels run as monomers either with or without an EGFP epitope tag. B, channels expressed with KChIP3 Western-blotted for Kv4.2 protein. Level of KChIP3 cDNA added is indicated to the right of the fractionation. As expected, Kv4.2 channels are still localized in the tetrameric fractions. Kv4.2(ϪZnB3) channels show a progressive shift to tetrameric fractions with increasing KChIP3 cDNA. An EGFP-tagged version of this mutant shows a similar shift to tetramer with expression of KChIP3 at levels 4ϫ greater than channel cDNA. C, channels co-expressed with KChIP3 Westernblotted to identify KChIP3 bound to the channel. Although the majority of the expressed KChIP3 is not assembled with channel at these expression levels, specific co-migration with channel in the tetrameric fractions is seen, indicating that KChIP3 integrates stably into normal and Zn 2ϩ site mutant channels. FIG. 5. Functional differences in Kv4.2(؊ZnB3) channels compared with wild-type channels expressed with KChIP3. A, activation curves in response to step depolarization from Ϫ30 mV to ϩ50 mV in 10-mV increments. Currents for Kv4.2(ϪZnB3), shown in red, and Kv4.2, shown in black, were normalized to the peak current at ϩ50 mV and plotted on the same time base. The Zn 2ϩ site mutant channels activate slightly less at the lower potentials and inactivate more rapidly at all potentials. An expanded time base inset compares the activation of the two channels at ϩ50 mV. As can be seen, the Zn 2ϩ site mutant channels appear to activate more rapidly, reaching a peak sooner than wild-type channels. B, steady state inactivation and peak activation properties of normal and Zn 2ϩ site mutant channels. Kv4.2(ϪZnB3) channels show a significant leftward shift in inactivation and a rightward shift in activation compared with wild-type channels. Statistical analyses were performed on the set of fits to individual oocytes using Student's t tests. Significant differences were found between mutant and wild type in the half-activation voltages, half-inactivation voltages, and activation slopes, but not inactivation slopes. take longer to reach a steady state for inactivation. Fig. 7D compares the kinetics for entry into the closed inactivated state for wild-type and Zn 2ϩ site mutant channels. We have plotted the average and standard deviations for the best single exponential fits to individual experiments at different potentials. At more negative potentials, the inactivation of Kv4.2(ϪZnB3) channels is significantly slower that wild-type Kv4.2 channels. Inactivation of the Kv4.2(ϪZnB3) channels however shows a much stronger voltage dependence than wild-type Kv4.2, and the traces cross over each other at approximately Ϫ52 mV. Above this potential, inactivation of Kv4.2(ϪZnB3) channels is more rapid, in agreement with our results on depolarized potentials shown in Fig. 6.
Recovery from Inactivation-The level of steady state inactivation is controlled by the ratio of the rate for entry into the inactivated state to the rate of recovery from the inactivated state. We therefore were interested in examining if changes in the recovery from inactivation, indicative of a more stable inactivated state, are significantly altered with the loss of the Zn 2ϩ site. Recovery was characterized using a typical two-pulse protocol where the time spent at the interpulse potential was varied to characterize the recovery time course. Fig. 8, A and B shows the recovery profiles for wild-type Kv4.2 and Kv4.2(ϪZnB3) channels. The amplitude of the second peak was measured and plotted versus the time spent at the interpulse potential for normal and ϪZnB channels in Fig. 8C. It is clear from these plots that the ϪZnB mutants recover more slowly from inactivation. For wild-type Kv4.2 expressed with KChIP3, the time constant to recover from inactivation at Ϫ80 mV is 81.7 Ϯ 4.1 ms; however, for Kv4.2(ϪZnB3) the FIG. 6. Differences in kinetic properties of wild-type Kv4.2 and Kv4.2(؊ZnB3) channels expressed with KChIP3 in response to step membrane depolarizations. A, measured time to peak for normal and Zn 2ϩ site mutant channel shows a significantly decreased time to peak for mutant channels at all potential above 0 mV. B, inactivation kinetics for wild type and ϪZnB3 mutant channels. Fastest inactivating current is Kv4.2 alone. Middle current is Kv4.2(ϪZnB3) expressed with KChIP3. Slowest inactivating current is wild type Kv4.2 expressed with KChIP3. Current decays were well fit with two exponentials when co-expressed with KChIP3, but a third faster exponential component is required to fit wild type expressed alone. Fits are overlaid on the traces and show a very good fit over the entire 1 time course for the current decay. C, measured time constants for the two exponential fits to the inactivation process. At all potentials 0 mV or above, the measured time constants are faster for Kv4.2(ϪZnB3) channels than for wild type. Three exponential fits for Kv4.2 expressed without KChIP3 at ϩ50 mV provided for comparison. D, the relative fraction of the inactivation process that is occurring through the fast inactivation process decreases at higher membrane potentials. The fraction of channels inactivating with the faster process is greater at all potentials for Kv4.2(ϪZnB3) channels than wild type. One-way ANOVAs were performed to compare the set of time constants measured for wild type versus mutant. Time to peak, slow and fast time constants of inactivation, as well as the fast time constant weight were found to be significantly different for mutant compared with wild type. Significant voltage dependence was found for time to peak and fast time constant weight for both mutant and wild type. Voltage dependence for the mutant fast and slow time constant was not significant. For wild-type channels, voltage dependence of the fast time constant was not significant, but the slow time constant was significantly voltage-dependent. recovery is dramatically slowed to 254.3 Ϯ 8.2 ms. In fact, this value is almost as slow as Kv4.2 expressed without KChIP. Therefore, loss of the Zn 2ϩ site eliminates much of the effect that KChIP3 produces to accelerate the recovery of Kv4.2 from inactivation. Fig. 8D summarizes the results from a range of interpulse potentials between Ϫ110 mV and Ϫ70 mV, showing that the slowing of recovery is consistent over this potential range.

DISCUSSION
In these studies, we have found that the KChIP3 auxiliary subunit protein has a dramatic ability to rescue the function of an assembly mutation of Kv4.2 subunit proteins. Kv4.2 subunits with the zinc site mutated are completely unable to form functional channels because of a loss of subunit tetramerization. Our results strongly suggest that this loss of function in these mutants is a specific effect related to the loss of Zn 2ϩ binding, rather than a nonspecific effect of these mutants: 1) All mutants to Zn 2ϩ coordination residues tested (single, double, triple, and quadruple) have very similar effects on channel assembly and have similar functional changes when rescued by KChIP3. 2) The four coordination residues mutated are located in 3 different secondary structural elements. 3) The coordination residue Cys 111 is found on the opposite surface and only joins the other three residues following folding and interaction of the assembly interface (9). The only clear relationship between these mutants is that they all disrupt Zn 2ϩ binding to this site.
KChIPs are revealed here to be active participants in the assembly of subunits, capable of binding to unassembled subunits and then driving them together to form functional channels. These results clearly provide important new insights into KChIP function. The mechanism by which KChIPs are able to drive channel assembly is not clear; however, previous results suggest some possible mechanisms. First, KChIP proteins have been shown to bind to the very N-terminal region of Kv4.2 subunits (12). So, one possible mechanism is that the floppy N-terminal chain folds back onto the T1 domain and allows the KChIP protein to interact FIG. 7. Comparison of the entry of wild-type and ؊ZnB3 mutant channel into the closed inactivated state at potentials near the V1 ⁄2 for steady state inactivation. A, inactivation of wild-type Kv4.2 channels at Ϫ55 mV is shown using a modified two pulse protocol. B, same pulse protocol for inactivation of Kv4.2(ϪZnB3) channels at Ϫ55 mV. Note the more complete inactivation of this mutant; however, the time course for reaching this final value is prolonged compared with wild type. C, plotted peak currents for the second pulse normalized by the first pulse for potentials bracketing the V1 ⁄2 for steady state inactivation. The enhanced inactivation for the same pulse potential is produced by the negative shift in V1 ⁄2 for Kv4.2(ϪZnB3). Comparing decays to approximately the same final value for Kv4.2 and Kv4.2(ϪZnB3), Ϫ50 mV, and Ϫ60 mV respectively, shows the prolonged decay time course for Kv4.2(ϪZnB3) closed state inactivation at the same final level of inactivation. D, summary data for closed state inactivation time course measured by single exponential fits to individual experiments. Kv4.2(ϪZnB3) shows much more voltage dependence in this time constant than wild-type Kv4.2 and crosses over to faster inactivation above approximately Ϫ52 mV. Significant differences were found in the voltage dependence as well as the time constants for inactivation at Ϫ55 and Ϫ60 mV.
with and rescue the assembly interface lost with the Zn 2ϩ site mutants. Such a possibility is suggested by yeast 2-hybrid experiments, which have shown that KChIP proteins can interact directly with specific regions of the T1 domain, providing a direct mechanism to modulate T1-T1 interactions (13). Indeed, single particle reconstruction experiments have suggested a binding of KChIPs along the T1 interaction interface (19). These models agree with our observations in SEC-FPLC studies that show a minimal impact on KChIP binding on the hydrodynamic radius of the channel, suggesting the KChIP protein is tightly integrated into the channel structure. The crystal structure of the KChIP-Kv4.2 interaction suggests another mechanism for driving assembly by the dimeric interactions between KChIPs bound to the Kv4.2 N terminus (12). This mechanism would not directly rescue the lost interactions within the T1 domain, but rather replace the energy lost with a complementary set of interactions that occur separately from the T1 domain. Future mutagenesis studies that can separate KChIP dimerization interactions from direct KChIP interactions with the T1 domain may allow us to distinguish between these different possibilities.
Another important question remaining to be answered is the relative importance of KChIPs driving subunit assembly compared with their trafficking properties in the formation of na-tive A-currents. Although it could be argued that there does not seem to be any problems in Kv4.2 subunit assembly without KChIPs, in native cells, where Kv4.2 expression is not being driven by a cytomegalovirus promoter, the situation could be very different. There, efficient assembly of Kv4 channels into a tetrameric channel may require auxiliary subunits to help the forming channel hold together long enough to find the required subunits out of the sea of thousands of other membrane proteins being synthesized at the same time.
Indeed, we may be creating a somewhat artificial distinction when we classify trafficking and assembly into two discrete processes. Examination of Fig. 4 shows that the subcellular distributions of wild-type and Zn 2ϩ site mutant channels are similar in the absence of KChIPs. Thus, cellular processes that trap the unassembled Zn 2ϩ site mutant channel proteins within the endoplasmic reticulum are similarly recognizing the wild-type protein as being improperly assembled until an interaction with KChIP completes the assembly and frees the protein to reach the cell surface. Biophysical studies on native neuronal A-currents and cardiac I to -currents have suggested that formation of functional currents in these native cells likely requires interactions with auxiliary subunits such as KChIP3 (24 -26). Of particular importance in determining if these processes represent distinct functional mechanisms will be to de- Channels were fully inactivated with a 1-s depolarization to ϩ50 mV, and then the channels were allowed to recover for varying lengths of time at different potentials. A, wild-type channel recovery at Ϫ80 mV, shows a rapid recovery of channel function. B, Kv4.2(ϪZnB3) mutant recovery at Ϫ80 mV is prolonged compared with wild type. C, average recovery from inactivation for normal and Kv4.2(ϪZnB3) mutant channels. The recovery curves were well fit with single exponentials. D, summary data for exponential fits to individual experiments at different recovery potentials. Recovery was ϳ3ϫ longer for Kv4.2(ϪZnB3) mutants compared with wild type at all potentials tested. Values at all potentials were significantly different between wild type and mutant.
termine whether mutations to KChIPs can be identified, which differentially affect their ability to rescue Zn 2ϩ site mutant Kv4 channel assembly without affecting wild-type channel trafficking functions, and visa versa.
Our results further show that an intact Zn 2ϩ binding site of Kv4.2 subunits is not absolutely required for channel function. For Kv4.2 channels, the most dramatic effect of Zn 2ϩ site mutations is a slower recovery from inactivation by 3-4-fold. At the negative potentials where this time constant is measured, its value is likely dominated by the rate constant for recovery from inactivation, and is thus a good indicator that the stability of the inactivated state of the channel has increased by Zn 2ϩ site mutation. In addition, at depolarized potentials, we find that the Zn 2ϩ site mutant channels peak sooner and inactivate more rapidly. Modeling studies have suggested that inactivation in Kv4 channels is caused by the channels accumulating in the closed inactivated state at both low and high voltages (21). The stabilization of the inactivated state by Zn 2ϩ site mutagenesis is only expected to accelerate the rate of inactivation at positive potentials if Kv4.2 channels significantly reopen from the closed inactivated state. An alternative explanation for the reduced time to peak and faster inactivation kinetics is that Zn 2ϩ site mutations also accelerate the entry into the inactivated state. The combination of these two effects would act together to left-shift the steady state inactivation curve for Zn 2ϩ site mutant channels.
Interestingly, Zn 2ϩ site mutants could accelerate inactivation by two different mechanisms, either accelerating the rate to enter the closed inactivated state, or increasing the fraction of activated channels that remain closed. Previous studies on Kv2.1 channels showed that chemical modification of N-terminal Cys residues prolonged the latency to first channel opening (27). Although the exact Cys residues responsible were not identified, the region implicated included the cysteine residues that form the T1 Zn 2ϩ site. If a similar effect on gating is occurring in Kv4.2 channels with Zn 2ϩ site mutations, this mechanism would enhance the rate of closed state inactivation of Kv4.2 channels. Although a prolonged time to first opening would seem to be contradictory to the reduced time to peak seen in Zn 2ϩ site mutant Kv4.2 channels, the accelerated inactivation of these channels would in effect truncate the activation process and create the appearance of faster activation. To further explore the relative importance of these different mechanisms will require detailed measurements of single channel behavior of normal and Zn 2ϩ site mutant channels.