Calmodulin Mediates Calcium-dependent Activation of the Intermediate Conductance KCa Channel,IKCa1 *

Small and intermediate conductance Ca2+-activated K+ channels play a crucial role in hyperpolarizing the membrane potential of excitable and nonexcitable cells. These channels are exquisitely sensitive to cytoplasmic Ca2+, yet their protein-coding regions do not contain consensus Ca2+-binding motifs. We investigated the involvement of an accessory protein in the Ca2+-dependent gating of hIKCa1, a human intermediate conductance channel expressed in peripheral tissues. Cal- modulin was found to interact strongly with the cytoplasmic carboxyl (C)-tail of hIKCa1 in a yeast two-hybrid system. Deletion analyses defined a requirement for the first 62 amino acids of the C-tail, and the binding of calmodulin to this region did not require Ca2+. The C-tail ofhSKCa3, a human neuronal small conductance channel, also bound calmodulin, whereas that of a voltage-gated K+channel, mKv1.3, did not. Calmodulin co-precipitated with the channel in cell lines transfected with hIKCa1, but not with mKv1.3-transfected lines. A mutant calmodulin, defective in Ca2+ sensing but retaining binding to the channel, dramatically reduced current amplitudes when co-expressed withhIKCa1 in mammalian cells. Co-expression with varying amounts of wild-type and mutant calmodulin resulted in a dominant-negative suppression of current, consistent with four calmodulin molecules being associated with the channel. Taken together, our results suggest that Ca2+-calmodulin-induced conformational changes in all four subunits are necessary for the channel to open.

Ca 2ϩ -mediated signaling events are central to the physiological activity of diverse cell types. Opening in response to changes in intracellular Ca 2ϩ ([Ca 2ϩ ] i ), Ca 2ϩ -activated K ϩ (K Ca ) 1 channels play an important role in modulating the Ca 2ϩ signaling cascade by regulating the membrane potential in both excitable and nonexcitable cells. Historically, these channels have been classified as large (BK Ca ), intermediate (IK Ca ), and small (SK Ca ) conductance channels based on their singlechannel conductance in symmetrical K ϩ solutions (1). BK Ca channels have a single channel conductance of 100 -250 pS, are opened by elevated [Ca 2ϩ ] i as well as by depolarization, and are blocked by the scorpion peptides charybdotoxin (ChTX) and iberiotoxin (2). These channels are abundant in smooth muscle and in neurons and are also present in other cells (2). BK Ca channels are composed of an ␣and a ␤-subunit. The ␣-subunit, encoded by the Slo gene (3)(4)(5), is a seven-transmembrane region protein with an extracellular N terminus (6). The ␤-subunit is a two-transmembrane region protein that, when associated with the channel, enhances the Ca 2ϩ sensing and toxin binding properties of the channel (7,8).
SK Ca channels have unitary conductances of 4 -14 pS; are highly sensitive to [Ca 2ϩ ] i , with activation in the 200 -500 nM range; and are voltage-independent (9,10). SK Ca channels are highly expressed in the central nervous system, where they modulate the firing pattern of neurons via the generation of slow membrane after-hyperpolarizations (10). SK Ca channels have also been described in skeletal muscle (11) and in human Jurkat T-cells (12). These channels are blocked by apamin, a peptide from bee venom, and by the scorpion peptide scyllatoxin (12)(13)(14). Three genes (SKCa1-3) within a novel subfamily encode SK Ca channels (13). SKCa1-3 gene products bear 70 -80% amino acid sequence identity to each other, and hydrophilicity analysis predicts that these proteins have six transmembrane helices with intracellular N and C termini (13,15). The hSKCa3 gene has recently been implicated in schizophrenia (15,16). IK Ca channels, unlike SK Ca channels, are predominantly expressed in peripheral tissues, including those of the hematopoietic system, colon, lung, placenta, and pancreas (17)(18)(19)(20)(21)(22)(23). These channels have intermediate single channel conductance values of 11-40 pS and can be pharmacologically distinguished from SK Ca channels by their sensitivity to block by ChTX and clotrimazole and by their insensitivity to apamin (20,22). Both SK Ca and IK Ca channels are voltage-independent and steeply sensitive to a rise in [Ca 2ϩ ] i . At least one gene encoding an IK Ca channel has been cloned from human and mouse tissues. Called IKCa1 (also called KCa4, SK4, and KCNN4), this gene has been shown to encode the native IK Ca channel in human T-lymphocytes (22,23) and erythrocytes (24 -27); some patients with Diamond-Blackfan anemia lack one allele of this gene (23). hIKCa1 shares little sequence identity with the Slo proteins, and only about 40% identity with the SKCa1-3 gene products. Thus, hIKCa1 constitutes a distinct subfamily within the extended K ϩ channel supergene family.
The Ca 2ϩ sensor for BK Ca channels resides in a negatively charged Ca 2ϩ bowl domain in the C-tail of the ␣-subunit (28,29). The ␤-subunit also contributes to the gating of these proteins (7). In marked contrast, the protein-coding regions of SKCa1-3 and hIKCa1 do not contain any EF-hand or Ca 2ϩ bowl motifs in their primary amino acid sequence, despite their exquisite Ca 2ϩ sensitivity. This observation led us to speculate that the Ca 2ϩ sensor for these channels either resides in a novel motif intrinsic to the channel or is provided by an accessory subunit that is tightly linked to channel activity. We investigated the latter possibility in a yeast two-hybrid system using hIKCa1 as our prototype. The Ca 2ϩ -binding protein calmodulin (CAM) was identified as a strong interacting partner of the C-tail of hIKCa1. Recently, CAM was shown to confer Ca 2ϩ sensitivity to SK Ca channel subfamily members (30). Here, we report that CAM binds to and is required for Ca 2ϩdependent activation of hIKCa1. Biochemical studies demonstrate that both hIKCa1 and hSKCa3 are prebound tightly to CAM in a Ca 2ϩ -independent fashion. Finally, we show by expression and patch-clamp recording that four CAMs are required to mediate the Ca 2ϩ -dependent channel activity of the hIKCa1 tetramer.

EXPERIMENTAL PROCEDURES
Clones, Mutants, and Vectors-We have previously reported the cloning of hIKCa1 (22,23), hSKCa3 (15), and mKv1.3 (31). Drosophila wild-type (WT) and mutant (B1234Q) CAMs with differing Ca 2ϩ sensitivities have been reported previously (32,33). The B1234Q mutant has all four EF-hands mutated; glutamates 31, 67, 104, and 140 are replaced by glutamine (33). PAGA2 vector was a kind gift of Lutz Birnbaumer (University of California, Los Angeles, CA). This vector is a pGEM3-based version of the pAGA vector, both of which contain the 5Ј-untranslated region of alfalfa virus RNA 4 and a 92-base pair poly(A) tail to increase stability of message and for efficient in vitro translation. The segments of DNA encoding the C-terminal tails of hIKCa1 (nucleotides 1252-1678; GenBank TM accession AF022797), hSKCa3 (nucleotides 1632-2193; GenBank TM accession number AF031815) and mKv1.3 (nucleotides 1736 -2112; GenBank TM accession number M30441) were subcloned into the PAGA2 vector using the polymerase chain reaction with engineered restriction sites. Both CAM clones were also subcloned into the PAGA2 vector. For co-precipitation and electrophysiology experiments (see below), the full-length hIKCa1 (Gen-Bank TM accession number AF033021) and mKv1.3 coding regions were fused in-frame with a N-terminal His 6 tag in the pcDNA3.1-His-C vector (Invitrogen, Carlsbad, CA). All clones were verified by sequencing.
Yeast Two-hybrid Screening-A 426-base pair fragment of hIKCa1 coding for residues 286 -427 in the cytoplasmic C-terminal tail of the channel was subcloned into the GAL4 DNA-binding vector (pAS2-1, CLONTECH, Palo Alto, CA) using polymerase chain reaction and engineered restriction sites. This construct was used as bait to screen an activated human leukocyte cDNA library (HL4021AB, CLONTECH). Screening procedures were performed according to the manufacturer's recommendations (CLONTECH PT3061-1). Several thousand putative positives were identified after first-round selection in growth medium; they were then subjected to the colony-lift lacZ assay. Positive blue colonies were sequenced using vector-specific primers.
Calmodulin Binding-Two methods were used to test for CAM binding to the channel proteins. The initial deletion constructs of hIKCa1 were generated by polymerase chain reaction as glutathione S-transferase (GST) fusions in the pGEX-6P-1 vector (Amersham Pharmacia Biotech), expressed in the Escherichia coli strain BL21-De3, and synthesis of the fusion proteins was induced with 0.1 mM isopropyl ␤-Dthiogalactoside in a liquid culture grown to A 600 of ϳ1.0. After 2.5 h at 37°C, cells were collected by centrifugation, resuspended in NETN lysis buffer (0.5% Nonidet P-40, 1 mM EDTA, 20 mM Tris-HCl (pH 8.0), 100 mM NaCl; 1.0 ml per 20 ml of culture) containing protease inhibitor mixture (complete protease inhibitor mixture tablets, Boehringer Mannheim), and lysed by sonication. The lysate was cleared by centrifugation at 10,000 ϫ g for 10 min at 4°C. GST fusion proteins in the supernatant were adsorbed for 30 min at room temperature to glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech) in NETN (1 volume of lysate:0.4 volume of 50% (v/v) slurry of Sepharose-GSH beads (Amersham Pharmacia Biotech) in NETN), which were then washed with binding buffer containing 1% (v/v) polyoxyethylene-9lauryl ether, 100 mM NaCl, 20 mM Tris-HCl, pH 8.0. For experiments investigating Ca 2ϩ dependence, the above buffer contained, in addition, 1 mM CaCl 2 or 2 mM EDTA. Slurries (50% (v/v)) of the bound Sepharose-GSH beads (Sepharose-GSH:GST fusions) were then incubated for 30 min at room temperature in 50 l of binding buffer containing [ 35 S]methionine-labeled hCAM, synthesized by coupled transcription-translation (TnT, Promega, Madison, WI) as described (34). The bound beads were washed three times with binding buffer and resuspended in 15 l (three volumes) of 2ϫ Laemmli's sample buffer. Proteins released from the beads by boiling in the presence of reducing reagent were analyzed by 4 -20% gradient SDS-PAGE followed by autoradiography to detect retention of hCAM by the channel-GST fusion proteins. To ensure equivalent protein loading, gels were stained with colloidal blue (Novex, San Diego, CA) to visualize the major protein band in each lane prior to autoradiography. Binding of WT-and B1234Q-CAMs to the C-tail of hIKCa1 was also determined using the GST pull-down method as above.
For all other experiments, channel constructs in the pAGA2 vector were radiolabeled with [ 35 S]methionine during coupled transcriptiontranslation using reagents from Promega. These constructs were incubated with CAM-Sepharose 4B beads (Amersham Pharmacia Biotech). Briefly, slurries of CAM beads (50% (v/v)) in binding buffer (as described above) were incubated with radiolabeled channel proteins that had been normalized for radioactive incorporation using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA). Equal cpms of each specific protein were added to 50 l of binding buffer including either 2 mM EDTA or 1 mM Ca 2ϩ . Binding and washing conditions were the same as for the GST-Sepharose experiments above. Proteins released from the beads by boiling in the presence of reducing reagent were analyzed by 18% SDS-PAGE followed by autoradiography.
His Tag Pull-down Assays-mKv1. 3 and hIKCa1 expression constructs in pcDNA-3.1His(C) were transfected into COS-7 cells using Fugene6 (Boehringer Manheim) according to the supplied protocol. About 40 h after transfection, 5 ϫ 10 6 cells were lysed in 10 mM HEPES (pH 7.4), 40 mM KCl, 0.75 mM EDTA (free Ca 2ϩ concentration, Ͻ1 nM), 1% Triton X-100, 10 mM ␤-mercaptoethanol, 0.25% deoxycholate, and protease inhibitors. After 20 min on ice, cells were Dounce-homogenized and centrifuged at 2900 ϫ g for 15 min to remove insoluble material. The soluble lysate was transferred to a clean tube and mixed with an equal volume of 2ϫ binding buffer (20 mM HEPES (pH 7.4), 200 mM KCl, 20% glycerol, 60 mM imidazole, 20 mM ␤-mercaptoethanol, and protease inhibitors). The diluted lysate containing the membrane fraction was incubated with Ni ϩ -NTA resin (Qiagen, Valencia, CA) for ϳ2 h at 4°C in order to immobilize the His-tagged channel protein. After extensively washing the resin with wash buffer (10 mM HEPES (pH 7.4), 100 mM KCl, 10% glycerol, 0.25% Triton X-100, 30 mM imidazole, 0.2 mM EDTA, 10 mM ␤-mercaptoethanol, and protease inhibitors), the channel protein was eluted with elution buffer (same as wash buffer but containing 400 mM imidazole). Proteins from the elution fraction, as well as from the flow-through, were separated by SDS-PAGE and transferred to polyvinylidene fluoride membranes. To determine whether CAM was preassociated with hIKCa1 or mKv1.3, a Western blot analysis was performed using an anti-CAM monoclonal antibody (Upstate Biotechnology, Lake Placid, NY).
Preparation of cRNA, Microinjection, and Whole Cell Recording-Rat basophilic leukemia (RBL) cells were maintained in a culture medium of Eagle's minimum essential medium (BIO-Whittaker, San Diego, CA) supplemented with 1 mM L-glutamine (Sigma) and 10% heat-inactivated fetal calf serum (Summit Biotechnology, Fort Collins, CO) and grown in a humidified, 5% CO 2 incubator at 37°C. Cells were plated to grow nonconfluently on glass 1 day prior to use for cRNA injection and electrophysiological experiments. T-lymphocytes were isolated from human peripheral blood and activated with phytohemagglutinin (DIFCO, Detroit, MI) as described previously (20). Prior to experimentation, cells were plated for 15 min on glass coverslips coated with poly-L-lysine (Sigma). For other experiments, we stably transfected the COS-7 cell line with hIKCa1; the biophysical properties of the hIKCa1 channels in these cells are indistinguishable from those of IK Ca channels in T-cells (data not shown). Plasmids containing the entire coding sequence of the hIKCa1 gene, WT-CAM, and B1234Q-CAM were linearized with NotI and in vitro transcribed with the T7 mMessage mMachine system (Ambion, Austin TX). Plasmids containing the mKv1.3 coding sequence were linearized with EcoRI and in vitro transcribed with the Sp6 version of the same kit. The resulting cRNA was phenol/chloroformpurified and stored at Ϫ75°C. RNA concentrations were determined to an accuracy of 25%, based on intensity of bands in agarose gel electrophoresis. The cRNA was diluted with fluorescein isothiocyanate-dextran (Sigma) (average M r , 10,000; 0.1% in 100 mM KCl). RBL cells were injected with an Eppendorf (Hamburg, Germany) microinjection system (Micro-manipulator 5171 and Transjector 5246) using injection capillaries (Femtotips®, Eppendorf) filled with the cRNA/fluorescein isothiocyanate solution, as described previously (35). Cells were visualized by fluorescence, and hIKCa1-specific currents were measured 4 -8 h after injection. Cells measured in the whole cell configuration were normally bathed in normal Ringer solution containing 160 mM NaCl, 4.5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, 10 mM glucose; adjusted to pH 7.4 with NaOH, with an osmolarity of 290 -320 mosM. In K ϩ Ringer solution, Na ϩ was replaced by K ϩ . A simple syringe-driven perfusion system was used to exchange the bath solutions in the recording chamber. The internal pipette solution with 1 M free Ca 2ϩ contained 145 mM K ϩ aspartate, 8.5 mM CaCl 2 , 2 mM MgCl 2 , 10 mM HEPES, 10 mM K 2 EGTA; adjusted to pH 7.2 with KOH, with an osmolarity of 290 -310 mosM. EGTA was omitted in the high Ca 2ϩ internal solution containing 1 mM CaCl 2 . Pipettes were pulled from glass capillaries, coated with Sylgard (Dow-Corning, Midland, MI), and fire-polished to resistances measured in the bath of 2-5 M⍀. Membrane currents were recorded with an EPC-9 patch-clamp amplifier (HEKA elektronik, Lambrecht, Germany) interfaced to a computer running acquisition and analysis software (Pulse and PulseFit; HEKA elektronik). Data were filtered at 1.5 kHz, and all voltages were corrected for a liquid junction potential offset of Ϫ13 mV for aspartate-based solutions. The holding potential in all experiments was Ϫ80 mV. For characterization of the hIKCa1 current, voltage ramp stimuli were used to assess channel activation by elevated [Ca 2ϩ ] i . RBL cells express an endogenous inwardly rectifying K ϩ channel that did not interfere with K Ca currents seen at depolarized potentials. Experiments were performed at room temperature (21-25°C). The CAM antagonists W7, trifluoperazine (TFP), and calmidazolium were purchased from Calbiochem (La Jolla, CA). ChTX was obtained either from Peptides International, (Louisville, KY) or from BACHEM Biosciences (King of Prussia, PA). Clotrimazole was purchased from Sigma. All slope conductances reported in the text are written as mean Ϯ S.E. (number of cells or experiments).

CAM Binds the C-tail of hIKCa1 in a Ca 2ϩ -independent
Manner-We searched for accessory molecules that bind to hIKCa1 using the yeast two-hybrid system. We reasoned that if such a molecule is involved in Ca 2ϩ sensing, it must be common to IK Ca and SK Ca channels, because their Ca 2ϩ sensitivities and gating behavior are remarkably similar (10,13,20,22,24). An amino acid alignment of hIKCa1 with hSKCa1, rSKCa2, and hSKCa3 revealed that except for the pore and the transmembrane regions, the proximal half of the cytoplasmic C-tail was the most highly conserved (Fig. 1); the C-tail of hIKCa1 was therefore employed as the bait. We chose an activated human leukocyte cDNA library to screen for interaction partners because hIKCa1 has been previously shown to be highly up-regulated in activated lymphocytes (20,22). A primary screen using the triple nutrient selection (Trp ϩ Leu ϩ His ϩ ) re-sulted in the identification of several thousand positive clones. A subsequent subscreen of 500 colonies yielded nine clones that were positive for ␤-galactosidase activity. Seven of these clones encoded CAM.
We next examined the ability of 35 S-labeled in vitro translated hIKCa1-C-tail to bind CAM-Sepharose as assessed by running the product on an SDS-PAGE gel followed by autoradiography. As shown in Fig. 2A, a radiolabeled band of ϳ28 kDa is visible (hIKCa1) consistent with the size of the hIKCa1 C-tail (Fig. 2B), indicating that hIKCa1 and CAM interact. CAM also binds to the C-tail of the SK Ca channel, hSKCa3 ( Fig.  2A, ϳ22-kDa band (hSKCa3)), but not the C-tail of the voltagegated K ϩ channel, mKv1.3 (Fig 2A, mKv1.3). As an additional specificity control, we performed GST pull-down experiments; CAM did not bind GST alone but interacted with the GST-hIKCa1 C-tail (Fig. 3A). Thus, CAM interacts specifically with members of the IK Ca and SK Ca family.
Surprisingly, the C-tail fragments of hIKCa1 and hSKCa3 bound CAM efficiently in buffers containing 2 mM EDTA and no added Ca 2ϩ ( Fig. 2A, hIKCa1 and mKv1.3, right panel). Deletion analysis revealed that the shorter 1-98 and 1-62 fragments of the hIKCa1 C-tail also bound CAM efficiently in Ca 2ϩ -free conditions (Fig. 3A, bottom panel; Fig. 3B 1. Amino acid alignment of the C-terminal tails of hIKCa1, rSKCa2, hSKCa3, and hSKCa1. Dashes represent identical residues, and asterisks correspond to conserved residues. Shaded areas represent identical or conserved residues. Numbering shown above the sequence alignment is based on the first amino acid in the C terminus (post-S6), as predicted by Kyte-Doolitle hydropathy plots. The numbers shown at the right of each sequence correspond to the actual positions of residues in the channel sequence. The C-tail of hIKCa1 corresponds to residues 286 -427 in the protein; the C-tail of hSKCa1 corresponds to residues 383-561 (GenBank TM accession U69883); the C-tail of rSKCa2 corresponds to residues 395-580 (GenBank TM accession number U69882); and the C-tail of hSKCa3 corresponds to residues 543-731.
1-72, and 1-82) bound CAM, but only in the presence of 1 mM Ca 2ϩ (Fig. 3A, top panel; Fig. 3B, left panel), whereas two others (1-50 and 93-142) did not bind CAM at all (Fig. 3). Thus, CAM interacts with the C-tails of hIKCa1 and hSKCa3 in the absence of Ca 2ϩ , and this property resides in a domain within the first 62 residues of the hIKCa1 C-tail. The segment between residues 62 and 82 appears to mask the Ca 2ϩ -independent interaction of CAM with hIKCa1, because the 1-72 and 1-82 fragments bind CAM only in the presence of Ca 2ϩ , whereas residues 82-98 appear to reverse the negative effect of 62-82. Removal of as yet unidentified motifs between residues 1 and 37 appears to unmask a Ca 2ϩ -dependent interaction with CAM. Interestingly, the 1-98 segment of the hIKCa1-C-tail, which contains the Ca 2ϩ -independent and Ca 2ϩ -dependent modulatory domains, shares a high degree of sequence similarity with the three members of the SK Ca family (Fig. 1).
CAM Co-precipitates with Full-length hIKCa1 in Transfected Cells-The binding data described above suggest that CAM is preassociated with the channel in cells with resting low [Ca 2ϩ ] i . If this were the case, it should be possible to co-precipitate CAM from cells expressing hIKCa1. To test this hypothesis, we expressed an N-terminal His-tagged fusion protein of hIKCa1 in COS-7 cells, prepared a crude membrane lysate in a Ca 2ϩfree solution, and passed the lysate through a nickel chelate column to allow the hIKCa1 channel to bind to the column via a His-nickel interaction. The column was washed extensively, and the unbound fraction was collected in the flow-through. The His-tagged hIKCa1 channel (along with any prebound accessory proteins) was then eluted with 400 mM imidazole. We examined the flow-through and the hIKCa1-containing eluate fraction for CAM using a anti-CAM monoclonal antibody. As negative controls, we used membrane lysates from untransfected cells and lysates from cells expressing a His-tagged version of the mKv1.3 channel that does not bind CAM ( Fig.  2A). As expected for a ubiquitous protein expressed at high levels in mammalian cells, CAM was detected in the flowthough fractions from untransfected cells (Fig. 4), mKv1.3transfected cells, and hIKCa4-transfected cells. In contrast, CAM was detected only in the hIKCa1-containing eluate, but not in the eluates from untransfected or mKv1.3-transfected cells (Fig. 4). Thus, CAM specifically co-precipitates with fulllength hIKCa1, but not mKv1.3, suggesting that the IK Ca channel is tightly bound to CAM under basal conditions in mammalian cells.
CAM Antagonists Do Not Alter K Ca Channel Function-In T lymphocytes or in mammalian cells expressing hIKCa1, elevating [Ca 2ϩ ] i rapidly opens IK Ca channels, revealing a voltageindependent K ϩ current with a reversal potential near Ϫ80 mV (20,22). To investigate the role of CAM in the function of IK Ca channels encoded by hIKCa1, we tested whether CAM antagonists might disrupt K Ca currents activated by dialysis of human T cells with a pipette solution containing 1 M [Ca 2ϩ ] i . Whole cell recordings revealed two components of K ϩ current, an immediately active voltage-gated K ϩ current encoded by hKv1.3, along with a rapidly activating IK Ca current (Fig. 5A,  traces 1 and 2), as reported previously (20). The time course of the slope conductance of this IK Ca current is shown in Fig. 5B. Treatment with the CAM antagonist W7 (10 M) had no effect on the IK Ca current at physiological potentials. Although it blocked both currents at depolarized potentials (Fig. 5A), this suppression is voltage-dependent and is thought to be mediated by a direct effect on the channel, rather than via CAM modulation (36). Another CAM antagonist, TFP (10 M), also had no effect on IK Ca currents when applied acutely; the slope conductance ratio, pre-TFP/post-TFP was 1.3 Ϯ 0.2 in six cells. Intact cells were also pretreated with TFP, W7, or 2 M calmidazolium for 15-30 min prior to recording, with no effect (not shown). Inclusion of 10 M W7 inside the patch pipette in combination with such pretreatment also had no effect; the slope conductance ratio in seven drug-treated cells relative to untreated cells was 1.3 Ϯ 0.3. Similar results were observed in COS-7 cells stably transfected with hIKCa1. We conclude that   (Figs. 2 and 4), as well as the inability of CAM antagonists to alter current through these channels (Fig. 5), supports the idea of a stable, nonconventional association between CAM and hIKCa1. Therefore, to study the interaction of these proteins, simultaneous new synthesis of each might be required. We injected combinations of cRNA encoding channel proteins plus cRNA encoding WT or mutant CAM into RBL cells, enabling us to investigate the effects and interactions of the resulting gene products using electrophysiological techniques. First, we characterized the physiological and pharmacological properties of hIKCa1 expressed after injection of the encoding cRNA alone. Robust currents exhibiting all the hallmarks of IK Ca channels were seen 4 -7 h postinjection (Fig. 6). The currents reversed near Ϫ80 mV in normal Ringer solution, and switching the bath solution to K ϩ -Ringer (164.5 mM K ϩ ) shifted the reversal potential to ϳ0 mV, as expected from the Nernst equation for a K ϩ -selective channel (not shown). ChTX or clotrimazole reduced the current in a dose-dependent manner with the expected potency for IK Ca channels in native tissue (20,22,24).
To determine whether CAM was responsible for the Ca 2ϩmediated gating of these channels, we exploited the availability of a Drosophila mutant (B1234Q) CAM. Drosophila and human CAM are identical at the amino acid level except at five positions. In B1234Q, glutamates (Glu 31 , Glu 67 , Glu 107 , and Glu 140 ) at the ϪZ coordination positions in each of the four Ca 2ϩbinding sites have been replaced by glutamine, resulting in a dramatically lower affinity for Ca 2ϩ (33). We reasoned that co-expressing hIKCa1 along with B1234Q would result in a significant reduction of current amplitudes. In order for this hypothesis to be tested by co-expression, it was first important to show that B1234Q bound hIKCa1 normally. The apo form of B1234Q is structurally similar to WT-CAM, the UV CD signal at 222 nm for the B1234Q-apo form being approximately 80% of the wild-type value (33). The apo form of B1234Q would therefore be expected to bind the hIKCa1 C-tail in a Ca 2ϩ -independent fashion. Consistent with this prediction, 35 S-labeled Drosophila WT-and B1234Q-CAM bound to GST-hIKCa1 C-tail both in the presence (Fig. 7, lanes 4 and 5) and absence (lanes 9 and 10) of Ca 2ϩ , as did 35 S-labeled hCAM (lanes 3 and 8). These CAMs did not bind GST alone in the presence or absence of Ca 2ϩ (Fig. 7, lanes 1, 2, 6, and 7).
Co-injection into RBL cells of WT-CAM cRNA together with FIG. 5. CAM antagonists do not inhibit hIKCa1. A, currents in activated T-cells were elicited by voltage ramps from Ϫ120 to 30 mV over 200 ms. At hyperpolarizing potentials, where only IK Ca currents are observed, acute application of 10 M W7 to the bath solution had no effect, whereas at depolarized potentials, the current was inhibited. Numbers on traces correspond to time of ramp (shown in B), where ramp 1 was taken immediately after break-in, ramp 2 at steady-state hIKCa1 current, and ramp 3 after ϳ6 min of exposure to 10 M W7. B, time course of hIKCa1 slope conductance activated by dialysis with 1 M free Ca 2ϩ , determined at Ϫ80 mV. After complete and stable activation of hIKCa1 currents, 10 M W7 was applied to the bath solution (indicated by the bar), but no effect on the K Ca current was observed. Ratio of mean slope conductance of K Ca 2 min after treatment relative to the same cell prior to treatment, 0.9 Ϯ 0.2 (six cells); same ratio for K V ϩ K Ca current at ϩ25 mV, 0.76 Ϯ 0.02 (six cells) cRNA encoding the hIKCa1 channel produced robust hIKCa1 currents in the whole cell mode with 1 M free Ca 2ϩ in the pipette. In marked contrast, co-injection of B1234Q cRNA with hIKCa1 cRNA resulted in an average 17-fold reduction in current amplitude. Fig. 8A shows the ramp currents obtained from individual cells. Aside from the endogenous inwardly rectifying K ϩ current, almost no additional current was seen in the B1234Q-containing cells, as compared with a large outward current observed in the cells with WT-CAM. From experiment to experiment, great variation was noted in the magnitude of currents seen in WT CAM-microinjected cells. However, on a given day, the currents observed in mutant or WT CAM-microinjected cells co-varied in a consistent manner, preserving a statistically significant difference in current ratio. Table I summarizes these results. Overall, injection of mutant CAM re-duced currents to 6% of WT CAM-microinjected cells. A reduction in the whole cell K V current was not observed when B1234Q was coinjected with mKv1.3 cRNA (mean: 427 pA for WT and 460 pA for B1234Q), ruling out the possibility that the effects on hIKCa1 currents were due to global inhibition of translation by B1234Q. The results of a typical experiment are shown in Fig. 8B and clearly demonstrate that B1234Q inhibits current through hIKCa1. B1234Q is incapable of the normal high affinity Ca 2ϩ binding and concomitant conformational changes seen in WT-CAM, but it does show a very limited conformational response that is completed upon attaining Ca 2ϩ levels of 1 mM (33). However, even with 1 mM Ca 2ϩ in the patch pipette, cells expressing B1234Q exhibited no appreciable current (data not shown). Thus, the minimal conformational changes in B1234Q are not adequate to gate the channel, and the normal conformational changes associated with Ca 2ϩ binding to CAM are essential for channel opening.
Dominant-negative Suppression of hIKCa1 by B1234Q-Like most K ϩ channels, hIKCa1 is anticipated to be tetrameric, but although each subunit could bind a single CAM, the Ca 2ϩ binding requirement for channel activation is unknown. It is conceivable that one functional CAM binding to Ca 2ϩ is sufficient to activate an hIKCa1 channel, or perhaps channel activity requires that all four subunits bind WT CAM and undergo the Ca 2ϩ -induced conformational change. To explore this question, we co-injected RBL cells with hIKCa1 cRNA and a mixture of B1234Q mutant and WT CAM. The mutant and WT CAM composition of hIKCa1 channels formed in such cells can be predicted by the binomial distribution. The proportion, P(r), of channels with r mutant subunits is given by the equation, where p is the fraction of mutant CAM in the CAM mix and n is the total number of subunits, four in the case of a K ϩ channel. The experimental results of microinjecting mutant and WT CAM RNA mixtures can then be compared with these predictions to determine the permitted number of CAM-binding subunits that most accurately represents the observed conductance. Because the mutant CAM appears totally unable to activate hIKCa1 (Fig. 8), channels with four B1234Q CAMs are presumed to be nonfunctional. Channels containing a mixture of mutant and WT CAM subunits would be expected to conduct only if hIKCa1 can be activated by fewer than four functional CAM-bound subunits. Thus, the current in cells microinjected with a mixture of mutant and WT CAM should reveal the number of functional CAMs required for active channels. Fig. 9 shows the K Ca current in cells with mixed mutant and WT CAM normalized to control K Ca currents in cells microinjected with WT CAM in parallel experiments performed on the same day. Table I summarizes the results of all experiments. For the case in which even a single mutant subunit will disrupt channel function, the binomial equation simplifies to the equation, as represented by the solid line (curve 0) in Fig. 9, with p plotted as the abscissa. If a single mutant subunit is tolerated in a functional channel, the proportion of conducting channels is shown by the equation, as depicted by Fig. 9, curve 1. Similarly expanded equations can be written for cases in which two or three mutant CAM are allowed. The data are well fitted only by the equation in which a single mutant subunit is sufficient to disrupt hIKCa1 func-FIG. 8. Co-injection of hIKCa1 with mutant CAM inhibits IK Ca currents. hIKCa1 cRNA was co-injected into RBL cells with either WT-CAM or B1234Q cRNA. Slope conductance was determined at potentials between -20 and Ϫ40 mV to avoid contamination with currents through the endogenous inward-rectifier channel present at potentials below -70 mV. A, ramp currents in RBL cells injected with hIKCa1 in combination with WT-CAM (WT) or hIKCa1 in combination with B1234Q (MUT). B, comparison of the slope conductance of uninjected cells and those co-injected with hIKCa1 or KV1.3 and either WT-CAM (WT) or B1234Q (MUT). Each circle represents the measurement of slope conductance of currents in a single cell approximately 2 min after establishing the whole cell mode. The bold lines illustrate the mean slope conductance for all cells in each column. The difference between the mean slope conductance of cells microinjected with B1234Q RNA and those microinjected with WT-CAM RNA was statistically significant, as illustrated by the one-tailed Student's t test (p Ͻ 0.02). Cells co-injected with KV1.3 and MUT or WT-CAM showed no significant difference in current at ϩ20 mV. tion. Hence, the presence of a B1234Q CAM is dominant-negative for the function of hIKCa1.

DISCUSSION
In the present study, we have demonstrated that CAM is prebound to the cytoplasmic C-tail of the intermediate conductance K Ca channel, hIKCa1, and mediates Ca 2ϩ -dependent gating of these channels. The first 98 amino acids in the C-tail of hIKCa1 contain subdomains that are critical for both Ca 2ϩ -dependent and Ca 2ϩ -independent CAM binding. Although this region contains several positively charged and hydrophobic residues reminiscent of CAM-binding sites (37,38), its lack of dependence on Ca 2ϩ for binding is noteworthy. We also show that known CAM antagonists W7 and TFP have no effect on hIKCa1 current, indicating a novel binding surface for the CAM-hIKCa1 interaction. Such Ca 2ϩ -independent binding of CAM to its target protein, although uncommon, has been reported for some molecules, including nitric oxide synthase, neurogranin, neuromodulin, phosphorylase kinase, and unconventional myosins (39).
Our results with an IK Ca channel parallel and complement those recently reported by Xia et al. (30) for SK Ca channels. In that study, CAM was shown to associate tightly with the Ctails of SK Ca channels in the absence of Ca 2ϩ . Co-expression in Xenopus oocytes of rSKCa2 and CAM mutants with lower Ca 2ϩ binding affinities resulted in a significant decrease in the Ca 2ϩ sensitivity of the expressed channel, thus providing the first evidence for a mechanism of Ca 2ϩ -gating by SK Ca channels.  (46). These helical rods cross at the bottom cytoplasmic surface (bundle crossing) and diverge at the extracellular surface to accommodate the P-regions. The upper end of the helices correspond to the S6 (M2 in KcsA) segments (black), and the region below the bundle crossing represents residues 1-10 of the hIKCa1 C-tail (white). Residues 11-62 of the critical C-tail CAM-binding region of each subunit are shown (in scale) as a separate single helix based on both Chou/Fasman and Robson algorithms for secondary structure predictions. Two CAM molecules (actual structure in the absence of Ca 2ϩ ) are shown in apposition to the C-tail; the other two subunits also associate with CAM. The spatial disposition of the CAMs does not imply interaction with the C-tail in any particular orientation.  Fig. 8 legend and are reported in nS as mean Ϯ S.E. (number of cells), with data taken approximately 2 min after establishing the whole cell configuration. Similar ratios of slope conductance were obtained for all experiments employing a given proportion of WT:MUT CAM RNA (see Fig. 9). The day-to-day variations in slope conductance resulted from injections of different amounts of hIKCal RNA.  Our finding of an identical mechanism for hIKCa1 expressed in mammalian cells confirms a common mechanism of Ca 2ϩ gating for both SK Ca and IK Ca channels, despite their ϳ40% overall sequence identity. Not surprisingly, the region in the C-tail of hIKCa1 that we identified as being critical for CAM binding (1-98) shows a high degree of sequence similarity with the corresponding regions in the three members of the SK Ca subfamily of K ϩ channels.
In addition to demonstrating that CAM can bind to and mediate the function of hIKCa1, we have demonstrated a strong suppression of IK Ca conductance when the channel is co-expressed with a mutant CAM with four defective EF-hand motifs (B1234Q CAM). The fact that the mutant CAM is so effective at competing with the endogenous protein for binding to the channel subunits suggests some mechanism for coassembly of newly synthesized CAM protein with the cytoplasmic tails of new channel molecules as they are folded on the endoplasmic reticulum membranes. Alternatively, it is possible that there is, in essence, no pool of "free" CAM to compete with the newly synthesized protein. At elevated [Ca 2ϩ ] i , this condition appears to apply, with free (non-target bound) CAM representing about 0.1% of total CAM protein (40). The situation at resting [Ca 2ϩ ] i is less well understood. However, studies in muscle cells suggest that the levels of CAM and its target proteins are carefully co-regulated even at resting levels of [Ca 2ϩ ] i (41).
If each subunit of hIKCa1 channel tetramer binds one molecule of Ca 2ϩ -free CAM, and if the concerted action of all four molecules is necessary for gating, then perturbing one interaction would impose a dominant-negative phenotype on the channel currents. Consistent with this hypothesis, the currents observed in cells microinjected with a 1:1 ratio of WT and mutant CAM cRNA along with the channel cRNA exhibited about 1 ⁄16th of the current magnitude observed in cells microinjected with channel cRNA and WT CAM alone. When the ratio of mutant to WT CAM was varied, we observed that the current relative to WT-CAM-microinjected cells agreed with the equation for disruption of channel activity by binding of a single mutant CAM (Fig. 9). Furthermore, even 1 mM Ca 2ϩ in the patch pipette, a concentration sufficient to elicit the conformational changes seen in B1234Q, did not rescue conductance when this mutant CAM was co-expressed with hIKCa1. These results imply that Ca 2ϩ -induced conformational changes must occur involving each prebound CAM in order to open the channel (Fig. 10). The requirement for four CAM molecules provides a structural basis for the previously determined steeply cooperative Ca 2ϩ dependence for activation of the lymphocyte K Ca channel encoded by hIKCa1 (Hill coefficients of 3-4) (20,42).
Our results point naturally toward a kinetic model for gating of the IK Ca channel. In the following scheme, C-2Ca 2ϩ ** 7 C-3Ca 2ϩ ** 7 C-4Ca 2ϩ ** 8 8 C-3Ca 2ϩ *** 7 C-4Ca 2ϩ *** 8 C-4Ca 2ϩ **** 8 OPEN SCHEME 1 C indicates closed channel conformations, and asterisks represent activated subunit conformations. Horizontal transitions represent Ca 2ϩ binding to CAM on the channel, and vertical transitions represent Ca 2ϩ -CAM-induced conformational changes in the channel subunit, with the number of asterisks symbolizing the number of activated subunits. In the above scheme, each subunit of a tetrameric channel is associated with a single CAM, which can bind up to four Ca 2ϩ ions to induce a conformational change in the channel subunit. For each of four independent subunits, we envision a sequential two-step activation process: first Ca 2ϩ binding by preassociated CAM, and then a conformational change in the IK Ca channel subunit to an activated conformation. In this scheme, it is imagined that conformational changes in the absence of Ca 2ϩ binding to CAM are so energetically unfavorable that the states with more activated subunits than Ca 2ϩ -CAM moieties do not exist, consistent with the fact that no IK Ca conductance is seen at low [Ca 2ϩ ] i . Such states, if they existed, would fill in the lower left-hand portion of the scheme. All four subunits must be activated before the channel opens. Although speculative, the kinetic diagram is similar to previous proposals for a variety of K Ca channels based upon single-channel data (43)(44)(45). These schemes predict that the steep Hill coefficient determined in functional measurements of the Ca 2ϩ sensitivity of channel opening arises from the requirement for Ca 2ϩ -induced conformational changes by each of four subunits in BK Ca , SK Ca , and, as proposed here, IK Ca channels. How might a conformational change in the C-tail of hIKCa1 result in opening of the pore? A comparison of the sequence of this region with that of the structurally defined bacterial potassium channel, KcsA, from Streptomyces lividans (46) suggests that the first 6 -10 residues of the hIKCa1 C-tail correspond to part of the inner helix that includes S6 (Fig. 10). More specifically, these residues represent the stretch of the inner helix lying below the "bundle crossing" (Fig. 10), and any Ca 2ϩ -CAM-induced conformational change in this segment could conceivably be transmitted along the helical rod, resulting in channel opening. Interestingly, recent studies on the voltagegated K ϩ channel, Shaker, suggest that gating occurs at the bundle crossing possibly due to conformational changes in this region (47). Two different algorithms (Chou/Fasman and Robson) predict that the remainder of the 1-62 segment has a high helical propensity, suggesting that the inner helix might extend further cytoplasmically (Fig. 10). Coupling of this segment with the inner helix might underlie calcium gating of hIKCa1. This heuristic model requires direct structural verification.
Although second messenger cascades involving CAM are known to modulate many ion channels (48), there is growing evidence of regulation by Ca 2ϩ -CAM through direct binding (49). These phenomena have been documented for the Paramecium Ca 2ϩ -activated sodium channels (50), the Drosophila Ca 2ϩ -permeable channels trp and trpl (51,52), the vertebrate photoreceptors and olfactory receptors involving cyclic nucleotide gated channels (53), the ryanodine receptor Ca 2ϩ -release channels (54), and the N-methyl-D-aspartate receptors (55). However, the region of the C-tail of hIKCa1 and hSKCa3 implicated in Ca 2ϩ -free CAM shows no obvious similarity to sequences with a comparable function in trpl (51) or the ryanodine receptor (56). In these examples, channel modulation involves either activation or deactivation by CAM. In contrast, the high affinity for Ca 2ϩ and the rapid activation kinetics of SK Ca and IK Ca channels demands a fast gating mechanism (45). This "near-intrinsic" requirement is provided by preassociated CAM molecules in a tight multimeric complex with the channel tetramer, converting a modest change in intracellular Ca 2ϩ to a quick, robust physiological response. Further biochemical, biophysical, and direct structural studies will help elucidate the mechanisms by which CAM-induced channel conformational changes in the C-tail translate into opening of IK Ca and SK Ca channels, leading to hyperpolarization and downstream signaling events.