Structural determinants of KvLQT1 control by the KCNE family of proteins.

KvLQT1 is a Shaker-like voltage-gated potassium channel that when complexed with minK (KCNE1) produces the slowly activating delayed rectifier I(ks). The emerging family of KCNE1-related peptides includes KCNE1 and KCNE3, both of which complex with KvLQT1 to produce functionally distinct currents. Namely I(ks), the slowly activating delayed rectifier current, is produced by KvLQT1/KCNE1, whereas KvLQT1/KCNE3 yields a more rapidly activating current with a distinct constitutively active component. We exploited these functional differences and the general structural similarities of KCNE1 and KCNE3 to study which physical regions are critical for control of KvLQT1 by making chimerical constructs of KCNE1 and KCNE3. By using this approach, we have found that a three-amino acid stretch within the transmembrane domain is necessary and sufficient to confer specificity of control of activation kinetics by KCNE1 and KCNE3. Moreover, chimera analysis showed that different regions within the transmembrane domain control deactivation rates. Our results help to provide a basis for understanding the mechanism by which KCNE proteins control K(+) channel activity.

genital sensorineural deafness. The nature of the interaction between KCNE1 and KvLQT1 has been the subject of much investigation. An extensive mutagenesis study of KCNE1 (12) originally uncovered a region of the protein within the Nterminal portion of the cytosolic domain that appeared critical for KCNE1 function; changes within this stretch were poorly tolerated. Romey et al. (13) provided biochemical evidence in support of an interaction between the pore region of KvLQT1 and the intracellular C terminus of KCNE1. Cysteine-scanning mutagenesis (14,15) provided evidence that residues in the transmembrane domain line the channel conduction pore. Tapper and George (16), using KCNE1 deletions and chimeras with a sodium channel ␤-subunit, showed that the KCNE1 C terminus alone cannot modify the KvLQT1 channel but rather that the presence of the KCNE1 transmembrane domain is also needed for this function.
Recently, a novel KCNE family member, KCNE3, was shown to interact with KvLQT1 to produce a potassium current with a constitutive component (2). This current resembles that seen in colonic epithelia, where both KCNE3 and KvLQT1 are expressed. Here we have exploited the marked differences in the regulation of KvLQT1 by these two proteins to investigate which regions of these two proteins mediate the differences in effects on KvLQT1. The approach of KCNE1/KCNE3 chimeras has the advantage of placing structural portions of KCNE1 in a homologous environment, but one that can modulate KvLQT1 function quite differently from KCNE1. By using this approach, we find that a three-amino acid segment within the transmembrane domain of KCNE1 is sufficient and necessary to determine the specificity of KCNE control of channel activation. Additionally, our results show a physical separation of determinants of activation and deactivation.

MATERIALS AND METHODS
Cloning of KCNE3-We detected human ESTs with homology to KCNE1 via TBLASTN search. By overlapping the ESTs we constructed the full-length cDNA. Forward and reverse primers from the 5Ј-and 3Ј-ends of the predicted coding region (5Ј-ATGGAGACTACCAATG-GAACGGAG and 3Ј-TTAGATCATAGACACACGGTTCTT) were used in 5Ј-and 3Ј-rapid amplification of cDNA ends PCR using the Marathon Ready cDNA kit (CLONTECH) and human stomach cDNA. Several rapid amplifications of cDNA end clones were sequenced and assembled to get the full-length cDNA. The sequence has been reported to Gen-Bank TM under accession number AF302494.
Construction of KCNE1/KCNE3 Chimeras-Chimeras were constructed using a PCR-based approach. Briefly, using PCR with Pfu turbo (Stratagene) and 5Ј-phosphorylated primers, we produced bluntended PCR fragments that corresponded to the regions of interest from KCNE1 and KCNE3. Blunt-ended fragments from KCNE1 were gelpurified, ligated overnight to a blunt-ended PCR product corresponding to the appropriate region of KCNE3, and re-amplified by PCR for subcloning into PCR Blunt II TOPO (Invitrogen). Chimera cDNAs were digested using HindIII and EcoRI and ligated into the mammalian expression plasmid pcDNA3 (Invitrogen). Nomenclature of the primers and chimeras corresponds to the order of fragments (M is minK (KCNE1) and K is KCNE3), followed by the splice site in human KCNE1 when the sequences are aligned as shown in Fig. 1b. Thus, a chimera corresponding to the N-terminal 47 residues of KCNE1 followed by KCNE3 residues for the remainder of the protein is named M(1-47)K. All primers used are in Fig. 1a. Primers below the shaded gray line are complementary to fragments of both KCNE1 and KCNE3 and were used to generate swaps of short (Ͻ7) amino acids.
Cell Culture and Transfection-CHO cells were maintained in Ham's F-12 media supplemented with 10% fetal calf serum and penicillin/ streptomycin at 37°C and 5% CO 2 . Gene transfer was performed using 15 g of Qiagen Midiprep purified plasmid DNA. Cells were electroporated at 225 V, 72 ⍀, 1800 microfarads with cytomix media (17). Cells were studied between 24 and 48 h after transfection. Molar ratios of 7:7:2 of KvLQT1:chimera:green fluorescent protein plasmid were used, to allow identification of transfected cells as previously described (18,19).
Electrophysiology-Cells were grown on sterile glass coverslips and placed in an acrylic/polystyrene perfusion chamber (Warner Instruments, Inc.) mounted on an inverted microscope outfitted with fluorescence optics and patch pipette micromanipulators. Extracellular solution was 150 mM NaCl, 1.8 mM CaCl 2 , 4 mM KCl, 1 mM MgCl 2 , 5 mM glucose, and 10 mM HEPES buffer, pH 7.4, at room temperature. Intracellular pipette solution was 126 mM KCl, 4 mM K-ATP, 2 mM MgSO 4 , 5 mM EGTA, 0.5 mM CaCl 2 , 25 mM HEPES buffer, pH 7.2, at room temperature. The whole cell configuration of the patch clamp technique was used to measure potassium currents (20). An Axopatch 1D patch clamp amplifier was used, and protocols were controlled via PC using pCLAMP8. Data were filtered using an 8-pole Bessel filter at 1 kHz and sampled at 4 kHz. For the protocol in Fig. 3c, lower trace, sampling was at 40 kHz. Activation curves are derived from tail currents at 0 mV following a series of 2 s. depolarizations and fitted to a Boltzmann curve using the equation, I ϭ 1/(1 ϩ exp((V h Ϫ V)/k)), where I is the measured tail current; V is the applied membrane voltage; V h is the voltage at half-maximal activation; and k is the slope factor. Whole cell current amplitudes were compared daily to a KCNE1/KvLQT1 control and normalized to this value to control for variations in transfection efficiency.

KCNE1 and KCNE3 Control KvLQT1
Activity- Fig. 2 shows whole cell current recordings from CHO cells transfected with KvLQT1 cDNA alone or in combination with either KCNE3 or KCNE1. As previously reported, coexpression with KCNE1 results in an increase in whole cell current amplitude (2.2-fold, see Table I) and the kinetics of both activation and deactivation are slowed (Fig. 2, a and b, and Table I) (4,5). Coexpression of KCNE3/KvLQT1, as previously reported (2), resulted in a current with a constitutively active component (Fig. 2c), as viewed on faster time scales and hyperpolarizations to Ϫ150 mV ( Fig.  2c, inset). We also observed a distinct time voltage-dependent component of activation and deactivation for the KvLQT1/ KCNE3 current (Fig. 2, c, e, and f). The voltage-dependent component had altered activation (KvLQT1, 24.2 Ϯ 2.3 ms, and KCNE3, 21.1 Ϯ 1.9 ms) and deactivation kinetics (KvLQT1, 39 Ϯ 1.1 ms, and KCNE3, 16 Ϯ 2.1 ms) with a shift in V h (KvLQT1, Ϫ15.2 Ϯ 0.52 mV, and KCNE3, 52.2 Ϯ 2.15 mV), and lacked obvious voltage-dependent inactivation (Fig. 2c), indicating that it too was a consequence of the association of Kv-LQT1 with KCNE3. Whole cell current amplitudes were also augmented relative to KvLQT1 alone, to a similar extent as KCNE1 (KCNE1, 2.13 Ϯ 0.14-fold, and KCNE3, 2.22 Ϯ 0.09fold, relative to KvLQT1, 1.00). Expression of KCNE3 alone did not yield currents (Fig. 2d).

KCNE Transmembrane Regions Determine Specificity of KCNE Control of KvLQT1
Kinetics-To determine the regions of KCNE1 and KCNE3 that are responsible for the different effects on KvLQT1 function, we swapped various segments of each to produce recombinant chimera cDNAs. We then coexpressed the chimerical cDNA with KvLQT1 for functional analysis of the resulting currents. These chimeras thus define the transmembrane region of KCNE1 from Val-48 to Leu-63 (corresponding to KCNE3 residues Leu-62 to Leu-77) that appears to mediate the specific effects of these proteins, a hypothesis that we confirmed with chimera KM(48 -63)K (Table I and Figs. 3e and 5).
Specificity of Control of KvLQT1 Activity Lies in a Small Region within the Transmembrane Segment-To examine the transmembrane region in finer detail, we subdivided it by constructing chimeras KM(48 -56)K that had a clear constitutive component present and activated with a rise time of 42.5 Ϯ 3.2 ms (Fig. 4c), and KM(48 -59)K (Fig. 4a) which, like KCNE1, showed no constitutive component and slow activation (rise time 691 Ϯ 34.9 ms). These chimeras showed that activation kinetics follow amino acids 57-59 in the KCNE1 transmembrane domain (KCNE3 amino acids 71-73). This result indicated that a region of only three residues was potentially responsible for the activation differences seen in KvLQT1-KCNE1 complexes. We confirmed this hypothesis by constructing three amino acid-swapped chimeras KM(57-59)K and MK(57-59)M. As seen in Fig. 4, d and e, and . A constitutively active current was detected for MK(57-59)M; however, it composed a smaller fraction of the total than that seen with wild-type KCNE3 (Table I), indicating that additional residues outside this region may be involved in generating the constitutive component.
Separation of KCNE Determinants of Activation and Deactivation Rates-In contrast to currents seen with KCNE1 or other I ks -like chimeras (Table I and (Table I). A possible explanation is that residues 48 -56 of KCNE1 confer slow deactivation, whereas in KCNE3, C-terminal residues (60 -63) control fast deactivation. When both segments are present in the same protein, however, the KCNE3 residues may act in a dominant fashion, but only when residues 57-59 are present as well (compare KM(48 -56)K and MK(57-63)M, Table I).
KCNE3 Induces a Constitutively Active K ϩ Current through Complex Physical Interaction with KvLQT1-We determined the presence of a constitutively active component to KCNE3 currents by looking for the appearance of an instantaneous current following depolarization, manifested as sharp corner before the onset of voltage-activated current (see Figs. 2c, 3, and 4). From the height of this step we estimated the fraction of total current that was attributable to the constitutive component (Table I, last column). The presence of constitutively active current was also seen using very fast sampling rates and membrane hyperpolarizations to eliminate the possibility of a very fast-activating voltage-dependent current (Figs. 2c and 4e). All chimeras containing KCNE3 amino acids 71-73 exhibited such a current. In contrast, chimera KM(57-59)K which consists essentially of KCNE3 with amino acids 71-73 replaced with KCNE1 residues 57-59 does not exhibit any observable constitutive component to its current, demonstrating that these residues are necessary for generating the constitutive component (Fig. 4, d and e) The fraction of total current provided by this constitutive component, however (0.18 Ϯ 0.03), was smaller for chimeras containing these and surrounding residues than for KCNE3 (constitutive component is 0.302 Ϯ 0.03 of total current). Thus, although residues 71-73 are necessary for constitutive activation, they are not sufficient for its full expression. Our results suggest that full control of the constitutively active component may be more complex than kinetics of voltage activation. Interestingly, Cd 2ϩ treatment of a T59C KCNE1 mutant (T58 in human KCNE1) leads to currents with a constitutively active component (14). The equivalent residue of KCNE3 is a valine, a substitution that like T59C abolishes the hydrogen bonding potential of the threonine hydroxyl group and introduces a larger residue in its place. DISCUSSION Our work has determined that small regions within the transmembrane domain of KCNE family members control the activation and deactivation rates of KvLQT1. Specifically, a three-mino acid stretch is sufficient and necessary to confer specificity of activation kinetics to KvLQT1, but neighboring regions appear to control the rates of deactivation. These same three amino acids also appear sufficient and necessary for at least partial generation of the constitutive component of current seen with KCNE3/KvLQT1 complexes.
Several groups have provided evidence that the KCNE1 C terminus is a region critical for its function (12,13,21). Most recently, Tapper and George (16) have shown that the C-terminal domain alone is not sufficient for slow activation when placed in a chimera with a sodium channel ␤-subunit and that the C-terminal half of the transmembrane domain was also necessary. Our work defines that portion of the transmembrane domain that is sufficient and necessary to confer functional specificity to KCNE1 and -3. Our results do not, however, rule out a role for the C terminus in KCNE1 function. The greatest homology between KCNE1 and KCNE3 resides in the previously determined critical C-terminal amino acids (Fig. 1). This region may be needed for anchoring, binding, or position-ing the transmembrane domain that we now identify as the effector. This would explain the results of earlier mutagenesis studies that showed the C-terminal region to be relatively intolerant to amino acid substitutions, as well as an analysis of KvLQT1 mutations that affect KvLQT1 function only in association with KCNE1 (22).
Our results indicate a separation of structural determinants of activation and deactivation kinetics. Thermodynamics can help elucidate the mechanism of KCNE control of channel activity by relating experimentally determined rates to energy states of the channel as it opens and closes. Changes in the voltage dependence of activation (V h ) are indicators of changes  , k, and current magnitudes were derived as described under "Materials and Methods." Activation rise time was determined as the time to 1 ⁄2 max following an 80 mV depolarization. Deactivation time constant was measured by fitting current at Ϫ120 mV following depolarization using a single exponential. The ratio of constitutive component to total current was determined by taking the height of the instantaneous rise in current following depolarization to 100 mV to the total steady-state current after voltage-activated current had reached a plateau. in the relative free energy of the closed and open states of the channel (⌬G indicates change in Gibbs' free energy). Rates of opening and closing of K channels, in contrast, reflect changes in the activation energy (⌬G ‡ ). This is the energy required to reach the transition state, the least stable intermediate conformation along the reaction coordinate of a molecule undergoing a physical change. KCNE1 and KCNE3 affect the rates of both opening and closing of the channel and the voltage dependence of activation. Our results provide a physical basis for each of these effects. That we observed nearly identical kinetics of activation between KCNE1 and chimera KM(57-59)K suggests that this region is necessary and sufficient to modulate the slow gating of activation seen with KvLQT1-KCNE1 complexes. We observed a ϳ30-mV difference in the V h between the two, however. This result indicates that the substitution of the KCNE1 residues 57-59 in KCNE3 was able to confer the KCNE1-like kinetics of activation of KvLQT1 (by increasing ⌬G ‡ ) independently of the V h (⌬G). The implication therefore is that KCNE1 residues 57-59 interact with the transition state of the channel during the closed 3 open transition but do so somewhat independently of the relative free energy of the closed and open states (⌬G). In addition, our results have dissected the region of KCNE1 controlling activation rate from that controlling deactivation rate. This result is not consistent with a view of KCNE1 acting simply as an "enzyme," modulating the activation energy of a reversible reaction. If that were true, we would expect the regions controlling activation and inactivation to be identical. Our observations imply that KCNE1 modulates the kinetics of KvLQT1 by interacting with different transition states during activation (the closed 3 open transition) and deactivation (the open 3 closed transition). While such a scenario is impossible in simple chemical equilibria, in the case of voltage-gated ion channels the two transitions take place under conditions of different electric field across the membrane. Channel structure is expected to be different during activation (depolarization) and deactivation (repolarization). Thus, the transition states for each process are also expected to be different.
Our data are consistent with a model in which the conformation of KvLQT1, when altered by the change in voltage across the membrane, presents different regions of the channel for interaction with KCNE1 and KCNE3. Following a depolarization, it appears that KCNE1 residues 57-59 (KCNE3 residues 71-73) are available to interact with a transition state to modulate the activation kinetics. Upon repolarization, the different conformations of KvLQT1 and KCNE1/3 allow other neighboring transmembrane regions of the proteins to interact. This interaction then modulates the stability of the reverse transition state (⌬G ‡ rev ). Our study has thus elucidated structural components of KCNE1 and KCNE3 that are necessary and sufficient to modulate the gating kinetics of KvLQT1 by interacting with several transition states. The technique of perturbing specific structural components by moving them into a homologous environment can serve to complement results from gating charge (23), crystallographic (24), and structural data (25,26). Further work and analysis of the interactions of KCNE family of proteins with K ϩ channel ␣-subunits may help to improve our understanding of gating mechanisms in a new way.