hSK4/hIK1, a Calmodulin-binding KCa Channel in Human T Lymphocytes

Human T lymphocytes express a Ca2+-activated K+ current (IK), whose roles and regulation are poorly understood. We amplified hSK4 cDNA from human T lymphoblasts, and we showed that its biophysical and pharmacological properties when stably expressed in Chinese hamster ovary cells were essentially identical to the native IK current. In activated lymphoblasts, hSK4 mRNA increased 14.6-fold (Kv1.3 mRNA increased 1.3-fold), with functional consequences. Proliferation was inhibited when Kv1.3 and IK were blocked in naive T cells, but IK block alone inhibited re-stimulated lymphoblasts. IK and Kv1.3 were involved in volume regulation, but IK was more important, particularly in lymphoblasts. hSK4 lacks known Ca2+-binding sites; however, we mapped a Ca2+-dependent calmodulin (CaM)-binding site to the proximal C terminus (Ct1) of hSK4. Full-length hSK4 produced a highly negative membrane potential (V m ) in Chinese hamster ovary cells, whereas the channels did not function when either Ct1 or the distal C terminus was deleted (V m ∼0 mV). Native IK (but not expressed hSK4) current was inhibited by CaM and CaM kinase antagonists at physiological V m values, suggesting modulation by an accessory molecule in native cells. Our results provide evidence for increased roles for IK/hSK4 in activated T cell functions; thus hSK4 may be a promising therapeutic target for disorders involving the secondary immune response.

Both voltage-gated and Ca 2ϩ -activated K ϩ (K Ca ) 1 channels are widely expressed in immune cells, including human T lymphocytes. Drugs that block voltage-gated Kv1.3 channels inhibit T lymphocyte activation and proliferation, volume regu-lation, and cell-mediated cytotoxicity (for review, see Ref. 1). Inasmuch as these functions involve Ca 2ϩ influx through channels activated by depletion of Ca 2ϩ stores, one widely proposed role for K ϩ channels is to maintain a negative membrane potential and large driving force for Ca 2ϩ entry. However, the relative roles of K Ca versus Kv1.3 channels in these cell functions are not known, partly owing to the previous lack of potent K Ca blockers that do not also block Kv1.3 channels.
Two K Ca channels have been found in lymphocytes and lymphocytic cell lines. They differ in biophysical and pharmacological properties (2)(3)(4)(5)(6). An apamin-sensitive, small conductance channel (7-8 pS) is the prevalent K Ca channel in the commonly used Jurkat T cell line (4) and is also present in rat T and human B lymphocytes (2,6). However, a corresponding apamin-sensitive whole-cell current has not been identified in normal human T cells, perhaps a result of channel rundown we observed after cell disruption (2). Instead, a K Ca channel we first described (2,3) is the prevalent K Ca channel in resting and activated human T lymphocytes. It is a charybdotoxin-sensitive, inwardly rectifying channel (15-35 pS in symmetrical K ϩ solutions (5,6)) that is commonly called "IK," for intermediate conductance K Ca . Recently, a molecular candidate for IK was cloned from a human placental cDNA library (hSK4 (7)) and subsequently from human pancreas (hIK1 (8)) and a human lymph node library (hKCa4 (9)).
IK current increases in the 3-4 days following activation of human T cells (5). Thus, it is anticipated that this K Ca current will be especially important for secondary immune responses of activated T cells (lymphoblasts), including proliferation and volume regulation. The regulatory volume decrease (RVD) that follows T cell swelling is known to depend on K ϩ (and Cl Ϫ ) channels (10 -12). Although Kv1.3 is involved in RVD in some resting T cells (13), the relative contribution of IK versus Kv1.3 channels is not known, either in resting human T cells or in lymphoblasts. We previously reported that intracellular Ca 2ϩ rises immediately after human T cells are exposed to a hypotonic shock (14); thus, we predicted that K Ca currents would also subserve RVD.
In the present study we cloned hSK4 from human T lymphoblasts, expressed the channels stably in CHO cells, and compared the salient biophysical and pharmacological properties of the native and cloned channels. All intrinsic properties examined were indistinguishable, supporting the view that hSK4 homotetramer forms the ␣ subunit of the IK channel of lymphoblasts. We found that hSK4 mRNA expression is strongly up-regulated after T cell activation; thus we predicted (and observed) an increased role for IK current in lymphoblasts compared with resting T cells. Although hSK4 is functionally a K Ca channel, that is activated by a rise in intracellular Ca 2ϩ , it was recently reported that brain SK channels are not gated directly by Ca 2ϩ but rather by Ca 2ϩ interacting with calmodulin that is irreversibly bound to the channel (15). We have presented preliminary data showing that the lymphoblast IK current is inhibited by antagonists of calmodulin and CaM kinase (16,17). We now show details of this inhibition of native IK current and that calmodulin binds directly to the hSK4 channel protein in a Ca 2ϩ -dependent manner. The CaM binding domain resides in the proximal part of the C terminus, since binding to this region occurs in the absence of flanking sequence and is eliminated in constructs lacking this region. Unlike the study of heterologously expressed brain SK channels (15), we provide evidence for additional modulation of hSK4 in lymphocytes, results that have important implications for the existence of accessory molecules and cell-specific K Ca channel regulation.

EXPERIMENTAL PROCEDURES
Cells-Human peripheral blood mononuclear cells were isolated on a Ficoll-Paque (Amersham Pharmacia Biotech, Baie D'Urfé, Quebec, Canada) density gradient. To purify resting T cells, monocytes and B lymphocytes were removed by adhering them to a nylon wool column for ϳ1 h and then eluting the T cells, which were then placed in culture medium containing RPMI 1640 with L-glutamine (Life Technologies, Inc., Burlington, Ontario, Canada) supplemented with 10% fetal bovine serum (Sigma, Mississauga, Ontario, Canada) and 50 g/ml gentamicin sulfate (Life Technologies, Inc.). Mitogenically activated T lymphocytes (lymphoblasts) were prepared by treating the mixed mononuclear cells with 7 g/ml phytohemagglutinin (PHA-P, Sigma) and growing them in culture medium in a humidified 37°C incubator with 5% CO 2 for 3-4 days. PHA selectively stimulates T cell proliferation, and monocytes die during prolonged culture.
Cloning and expression of hSK4 in CHO cells was as described previously (7). The full-length hSK4 clone was isolated from a human placental cDNA library, subcloned into pcDNA3 (Invitrogen, Carlsbad, CA), and transfected, using LipofectAMINE (Life Technologies, Inc.), into Chinese hamster ovary (CHO) cells that had been grown to ϳ50% confluency. Stable transfectants were selected using Geneticin (Life Technologies, Inc.) and subsequently separated into clonal populations by single-cell sorting (FACSIV, Yale Cell-sorting Facility). Two cell lines with similar current densities were selected for electrophysiological characterization. These were grown in Iscove's modified Dulbecco's medium, supplemented with 10% fetal bovine serum, HT supplement, antibiotic/antimycotic, and 1 mg/ml Geneticin (all reagents from Life Technologies, Inc.).
Preparing RNA Probes-Total RNA was isolated from resting and activated human T lymphocytes, rat lung, and human placenta using the guanidinium isothiocyanate method (18) and subjected to DNase I digestion (0.1 units/ml, 15 min, 37°C; Amersham Pharmacia Biotech, Toronto, Ontario, Canada) to eliminate genomic contamination. First strand cDNA was synthesized according to the manufacturer's instructions (Amersham Pharmacia Biotech) using an oligo(dT)-based primer. The cDNA was then used as a template for PCR reactions using the following primers: Kv1.3 (GenBank TM accession number M30312) forward primer 5Ј AATGAGTACTTCTTCGACCGCAACAGACCCAGCTT-CGA 3Ј and reverse primer 5Ј CCAATGAAAAGGAAAATGAGCAGCC-CCAG 3Ј; hSK4 (GenBank TM accession number AF000972) forward primer 5Ј GTGCGTGCAGGATTTAGG 3Ј and reverse primer 5Ј TGCT-AAGCAGCTCAGTCAGGG 3Ј. The PCR reaction was conducted with 1.5 mM MgCl 2 , 0.5 M forward and reverse primers, and 10% of the cDNA reaction mixture, using a Minicycler system (MJ Research, Watertown, MA). After incubating the mixture at 85°C for 1 min, 1.25 units of Taq DNA polymerase (Sangon Ltd., Toronto, Ontario, Canada) was added and heated to 94°C for 3 min, followed by 30 cycles through a 1-min denaturation step at 94°C, a 1-min annealing step at 50°C, and a 3-min extension step at 72°C. A final extension for 5 min at 72°C was followed by incubation at 4°C until further processing of samples. After PCR, the DNA products were resolved in 2% agarose gels containing 0.5 mg/ml ethidium bromide. The identities of PCR-amplified fragments of the predicted sizes (856 bp for hSK4 and 790 bp for Kv1.3) were confirmed by restriction endonuclease digestion, which yielded the correct size bands on an agarose gel. By using additional hSK4 PCR primers, we then amplified cDNA encoding the entire open reading frame (1284 bp). Cloning and subsequent sequencing revealed 100% homology with the published amino acid sequences of hSK4/hIK1/hKCa4 (7)(8)(9).
Ribonuclease Protection Assays-RNase protection assays on T cells, lung, and placenta were performed using the amplified cDNA fragments for hSK4 and Kv1.3 as probes. To obtain RNA probes, these fragments were linearized with SmaI (for Kv1.3) or MluI (for hSK4) and in vitro transcribed with [␣-32 P]dUTP and T7 RNA polymerase, yielding RNA transcripts of 579 (Kv1.3) and 254 bp (hSK4). However, since about 66 bp of non-hybridizing vector sequence was included in each probe, the protected channel fragments should be 513 (Kv1.3) and 188 bp (hSK4). A control plasmid (mouse ␤-actin cDNA), purchased from Ambion and transcribed according to the supplier's instructions, should yield an RNase-protected fragment of 250 bp. Because of the great abundance of ␤-actin mRNA in these cells, the ␤-actin probe was labeled to a low specific activity (ϳ600-fold lower than hSK4 and Kv1.3) to allow simultaneous quantitation. For each experimental data point, 10 g of total RNA was used, with 5 g of yeast tRNA as a negative control for probe self-protection bands. Following hybridization and RNase digestion, the samples were electrophoresed in polyacrylamide gels, dried, and exposed overnight to x-ray film (X-Omat, Eastman Kodak Co.). Specific signals were quantified by densitometic analysis of the developed film using a Bio-Rad model Gs-670 densitometer, and results are expressed as mean Ϯ S.D. with statistical analyses using the Student's t test.
hSK4-Flag DNA Constructs-An XhoI site was added by PCR to the C terminus of the full-length hSK4, and then a pair of Flag-encoding oligonucleotides was spliced into this site. This construct was used to create a second proto-construct in which all transmembrane domains were eliminated, and a consensus Kozak sequence was added to the beginning of the cytoplasmic C terminus. Morph mutagenesis (5 Prime 3 3 Prime, Inc., Arapaho, CO) was used to add a second, silent EcoRV site and a second XhoI site to the C termini of each proto-construct. Derivatives of these proto-constructs (for details, see text accompanying Fig. 7) were created by cutting out the fragments flanked by either EcoRV or XhoI and religating the larger fragment.
Calmodulin Affinity Chromatography-CHO cells stably expressing Flag-tagged hSK4 constructs were grown to confluency in 100-mm Petri dishes. To assess binding of hSK4 protein to calmodulin-Sepharose, we used methods modified from Chapin et al. (19). The dishes were washed three times in cold phosphate-buffered saline containing calcium and magnesium. Then, 1 ml of ice-cold solubilization buffer was added, which contained 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 1% Triton X-114 (Sigma), and protease inhibitors (5 g/ml pepstatin, 10 g/ml chymostatin, 5 g/ml leupeptin, 10 g/ml antipain, 500 M benzamidine, 0.1% Trasylol), and the dishes were rotated for 15 min at 4°C. The lysate was triturated with a 23.5-gauge needle, and after removing insoluble material by centrifugation (15 min, 4°C) the supernatant was transferred to a new tube to which 6 l of 0.5 M EGTA was added (final concentration, 3 mM) to chelate the calcium. To promote Triton X-114 phase partitioning, this solution was warmed to 37°C for 3 min and then centrifuged (5 min, room temperature) at full speed in a microcentrifuge. The detergent phase (bottom) was resuspended (5 min, 4°C) in 900 g of cold calcium-free wash buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 3 mM EGTA). The entire phase-separation procedure was repeated, and then the detergent phase was diluted 9:1 in an ice-cold "binding" buffer containing calcium (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 0.5 mM CaCl 2 ) and incubated for 5 min at 4°C. The protein samples were pre-cleared (1-3 h, 4°C) with 50 l of Sepharose Cl-2B.
The pre-cleared sample (500 l) plus binding buffer (450 l) was added to a 20% slurry (50 l) of calmodulin-conjugated agarose beads (Sigma) that had been equilibrated through a series of buffers, ending with binding buffer. The entire sample in binding buffer was incubated overnight at 4°C on a rotator. For CaM antagonist competition experiments, 1000 nM calmidazolium was added to the sample slurry before the overnight incubation. The beads were pelleted by a brief microcentrifuge spin (ϳ5 s, 4°C) and washed with 1 ml of binding buffer containing the detergent, Triton X-100, first at 0.5% and then at 0.05%. This was followed by a 5-min incubation at 4°C in the 0.05% Triton X-100 solution, with or without 3 mM EGTA. The beads were harvested as before and washed once with detergent-free binding buffer with or without EGTA. The proteins were dissociated from the calmodulinconjugated agarose beads by boiling them 5 min in 60 l of Laemmli buffer. The resulting protein sample was analyzed by SDS-polyacrylamide gel electrophoresis and Western blotting using a biotinylated anti-Flag antibody and an avidin-horseradish peroxidase-coupled secondary antibody (both from Sigma).
Patch Clamp Electrophysiology, T Cells-Activated T lymphoblasts were used 3-4 days after PHA stimulation, at which time IK current amplitudes are much larger than in resting T cells at the same cytoplasmic Ca 2ϩ concentration (5). Whole-cell currents were measured using an Axopatch 200 amplifier, with 8 -12 M⍀ pipettes. During data acquisition, capacitive currents were canceled by analogue subtraction, 50 -70% series resistance compensation was used, and all currents were filtered at 2 kHz. All voltages were corrected to account for the junction potential between bath and pipette solutions. The bath contained the following (in mM): 145 sodium aspartate, 5 KCl, 1 MgCl 2 , 1 CaCl 2 , 5 HEPES, adjusted with NaOH to pH 7.4, 270 -283 mosmol. The K ϩ selectivity was verified and current rectification examined by replacing the sodium aspartate with 80 mM sodium aspartate, 65 mM potassium aspartate (70 mM K ϩ solution), or 145 mM potassium aspartate (150 K ϩ solution). To activate fully the IK channels (5), we used a high Ca 2ϩ (1.1 M) pipette solution, consisting of the following (in mM): 140 potassium aspartate, 1 K 4 BAPTA, 2 K 2 ATP, 0.9 CaCl 2 , 1 MgCl 2 , 5 HEPES adjusted to pH 7.2 with KOH, 260 -270 mosmol. Contributions from volume-sensitive Cl Ϫ currents (20) were small since aspartate was used in the bath and pipette, and the internal solution was slightly hypoosmotic. Fresh K 2 ATP (Sigma) was always added to pipette solutions just before use to help maintain channel and second-messenger activity during whole-cell recording.
Free Ca 2ϩ concentrations were calculated assuming a dissociation constant (K d ) of 10 Ϫ11 for the EGTA 4-⅐2 Ca 2ϩ complex and 10 Ϫ7 M for the BAPTA 4-⅐2 Ca 2ϩ complex at pH 7.2, and allowing for the weak calcium binding by ATP. When calmodulin (CaM) was added to the pipette, CaCl 2 was increased to 0.91 mM (rather than 0.90 mM) to maintain the free Ca 2ϩ concentration (Ca 2ϩ i ) at 1.1 M, according to the following chemical information. Ca 2ϩ binds to two globular domains in CaM, each with two Ca 2ϩ -binding sites (E-F hands) for a total of four binding sites (21). In principle, CaM can reduce free Ca 2ϩ i , but the effect of adding 50 M CaM will be very small, as follows. At physiological KCl concentrations (K ϩ competes with Ca 2ϩ binding to some extent), the affinity of CaM for Ca 2ϩ is ϳ100-fold lower than that of BAPTA, which is present at 1 mM in our pipette solutions. Since the mean of log K d is 5 for CaM (22) versus 7 for BAPTA, the effect of 50 M CaM on free Ca 2ϩ i is equivalent to adding a further 0.5 M BAPTA.
Data analysis was performed using pCLAMP (version 5.5 or 6, Axon Instruments, Foster City, CA), SigmaPlot (version 2, Jandel, San Rafael, CA), and Origin (version 4.1, Microcal, Northampton, MA). Where appropriate, data are presented as mean Ϯ S.E., with paired Student's t tests (when each cell acted as its own control) or unpaired t tests used to determine the statistical significance of differences (95% confidence interval). All recordings were made at room temperature (19 -22°C), except when KN-62 and KN-04 were used at 37°C.
Cell Proliferation-Activated lymphoblasts (stimulated for 3 days) and resting (naive) T cells were prepared as above and then seeded at 2 ϫ 10 4 cells/well in 96-well plates (Corning Glass, Corning, NY) in culture medium. Cells were first incubated with or without K ϩ channel blockers for 10 min, and then 7 g/ml PHA-P was added to either initiate or re-stimulate proliferation. After a further 3 days growth, the plates were spun (1,200 rpm for 15 min at room temperature); the supernatant was removed, and the plates were frozen at Ϫ80°C (to aid subsequent cell lysis). Proliferation was calculated from the change in total nucleic acid content in each well using an assay (CyQUANT kit, Molecular Probes, Eugene, OR) that measures the signal generated by binding of a fluorescent dye to nucleic acids, a signal that increases in proportion to the number of cells. Fluorescence signals were read from a 96-well plate using a plate-reader (CytoFluor II, PerSeptive Biosystems Inc., Framingham, MA) with excitation at 485 nm and emission at 530 nm. For each treatment, the average fluorescence from control wells (lymphocytes without mitogen or channel blocker) was subtracted from the average final fluorescence from treated wells to yield the change in fluorescence (nucleic acid content) during 3 days of proliferation. Results are expressed as the mean increase in fluorescence Ϯ S.D. (n ϭ 4 experiments, 4 replicates/experiment), and a Bonferroni multiple comparison test was used for statistical analysis.
Flow Cytometric Analysis of Regulatory Volume Decrease (RVD)-Changes in cell volume were measured using a flow cytometer (FACScan, Becton Dickinson, CA) by monitoring changes in right angle light scattering (side scatter) as an index of cell volume (23). To indicate percent change from the initial volume, we calculated a swelling index as (1/SSC h )/(1/SSC i ) ϫ 100, where SSC h is the average side scatter at each time after exposure to hypotonic medium and SSC i is the average value in isotonic medium. The flow cytometer was set to exclude dead cells and debris by omitting cells that stained with propidium iodide (0.04 g/ml), a nuclear marker excluded by live cells.
For control volume measurements, resting or activated T cells were suspended in isotonic medium which contained the following (in mM): 140 NaCl, 5 KCl, 1 MgCl 2 , 1 CaCl 2 , 10 HEPES, pH 7.4, 285 mosmol. Initial measurements were taken from 5000 live cells, and then aliquots of the same cells were exposed to a hypotonic medium (56% of normal osmolarity) with or without K ϩ channel blockers. The hypotonic medium contained the following (in mM): 70 NaCl, 5 KCl, 1 MgCl 2 , 1 CaCl 2 , 10 HEPES, pH 7.4, 159 mosmol. Side scatter was recorded from 5000 live cells every 30 s for the first 3 min and then at 6 min after the hypotonic shock. Whenever K ϩ channel blockers were used, the initial volume was measured in the presence of drug before exposing the cells to hypotonic medium. Data were analyzed using CellQuest software (version 3.0.1f, Becton Dickinson). Values are presented as mean Ϯ S.D. of at least 4 experiments per treatment, and a Bonferroni multiple comparison test was used for statistical analysis. Experiments were performed at room temperature (21-23°C).
Chemicals-We used the K ϩ channel blockers, charybdotoxin and iberiotoxin (purchased from Peptides International, Louisville, KY, or a gift from V. Gribkoff and S. Dworetzky, Bristol-Myers Squibb Co.), margatoxin and agitoxin-2 (Alomone Laboratories, Jerusalem, Israel), and apamin, d-tubocurarine, and clotrimazole (Sigma). The calmodulin antagonists, trifluoperazine and W-7 (N-(6-aminohexyl)-5-chloro-1-sulfonamide) were from Sigma. Bovine brain calmodulin, rat brain CaMKII, calmidazolium (compound R24571), and KN-62 an analogue of KN-62 without CaM kinase inhibitor activity, was purchased from Seikagaku America (Rockville, MD). These reagents (except KN-62 and KN-04) were prepared in bath saline, frozen in aliquots, and thawed just before use. KN-62 and KN-04 stock solutions were prepared in Me 2 SO and diluted in bath saline before use. The Me 2 SO concentration in the final bathing solution was Յ0.5% for T cells and Յ1% for CHO cells. Whenever cells were treated with KN-62, control cells were incubated in the same concentration of Me 2 SO for the same duration.

RESULTS
Up-regulation of hSK4 and Kv1.3 mRNA Expression in Activated T Lymphocytes-We amplified DNA corresponding to full-length transcripts of hSK4 and Kv1.3 channels from activated human T lymphocytes. The open reading frame of the full-length hSK4 clone was 1284 bp, which encodes a protein of 428 amino acids that is 100% identical to the recently cloned hSK4 (7), hIK1 (8), and hKCa4 (9). RNase protection assays ( Fig. 1, A-C) on total RNA from resting and activated human T cells were used to determine if changes in mRNA expression correlate with previously observed increases in amplitude of the two K ϩ currents in lymphoblasts. Human placenta and rat lung were used as positive controls, and yeast tRNA was used as a negative control. In lymphoblasts, there was a 14.6-fold increase in hSK4 mRNA (p Ͻ 0.001, n ϭ 4) and a 1.3-fold increase in Kv1.3 mRNA (p Ͻ 0.001, n ϭ 4) 3-4 days after mitogenic stimulation.
Biophysical and Pharmacological Properties of the Lymphoblast IK Are the Same as hSK4 Expressed in CHO Cells-Charybdotoxin (ChTx) is an effective blocker of native IK currents in lymphocytes (3,5) and of expressed hSK4/hIK1/hKCa4 channels (7)(8)(9). In lymphoblasts, with free Ca 2ϩ in the pipette buffered to 1.1 M to maximally activate IK, typical current responses to voltage steps or ramps were resolved into the following three components: Kv1.3, IK, and an anion current (Fig. 1D). Separation of the currents was achieved by first blocking Kv1.3 with margatoxin (MgTx, 2.5 nM) and then add-ing ChTx (15 nM) to block IK, leaving the anion current (Cl Ϫ Nernst potential about Ϫ22 mV) which we have previously characterized (20). For subsequent IK measurements we blocked Kv1.3 with 2.5-5 nM MgTx, measured the remaining current, and then blocked IK with 10 -20 nM ChTx and subtracted the remaining anion current. (This procedure was not necessary for stably transfected CHO cells, wherein endogenous currents were negligible compared with the large hSK4 currents.) The ChTx-sensitive currents ( Fig. 2A) were K ϩ -selective, as seen from the intersection of their current versus voltage (I-V) curves with the anion current (at Ϫ78 mV; Nernst potential, E K ϭ Ϫ86 mV). We further confirmed their K ϩ selectivity using high K ϩ bathing solutions, wherein the reversal potentials of both Kv1.3 and IK were commensurate with changes in the external K ϩ concentration: at 70 mM, reversal potential, E rev ϭ Ϫ15 to Ϫ17 mV, and at 150 mM E rev ϭ Ϫ2 to 0 mV (after junction potential corrections). Expressed hSK4 and native IK currents were not time-or voltage-dependent, even during long voltage clamp steps (Figs. 1D and 2, B and C) (5). Their wholecell I-V relations ( Fig. 2A) were nearly linear over a wide voltage range under physiological Na ϩ /K ϩ gradients (also Fig.  1D) but rectified inwardly with high K ϩ concentrations in the bath and pipette. These biophysical features are consistent with previous studies of native IK in lymphocytes (5,24), and inward rectification in high K ϩ was observed for the channels expressed in Xenopus oocytes (hIK1 (8)) and in HEK cells (hKCa4 (9)). For the remaining experiments, whole-cell recordings from both cell types were made with high concentrations of external Na ϩ and internal K ϩ .
For comparison with the literature we tested ChTx (Fig. 2, B and D) and found it to be a potent blocker of hSK4 in CHO cells (IC 50 ϭ 1.7 nM, n ϭ 7) and of IK in T lymphoblasts (IC 50 ϭ 6 nM, n ϭ 3). For expressed hSK4 the dose dependence was calculated from the initial block during each voltage clamp step, to avoid the time-and voltage-dependent relief of block at depolarized potentials (Fig. 2B). Clotrimazole blocks Ca 2ϩ -activated K ϩ fluxes in thymocytes and red blood cells (IC 50  Activated T cells were used on days 3-4 after mitogenic stimulation. Total RNA (10 g) from the indicated cells and tissues was hybridized to 6 ϫ 10 4 cpm of 32 P-labeled hSK4 or Kv1.3 probe and 1 ϫ 10 3 cpm of ␤ actin probe. Left-hand panels show the full-length probe and self-protected bands determined using 5 g of yeast tRNA as a negative control. See "Experimental Procedures" for details and predicted band sizes. C, histogram showing substantial up-regulation of hSK4 mRNA (left-hand bars) in activated T lymphocytes (14.6-fold, p Ͻ 0.001) and a small increase in Kv1.3 mRNA (right-hand bars, 1.3-fold, p Ͻ 0.001). Values are densitometer readings (mean Ϯ S.E., four independent mRNA samples) normalized to ␤-actin counts in the same samples. D, co-existing IK and Kv1.3 currents in activated human T lymphocytes: typical currents for a cell 3 days after PHA stimulation. The pipette solution was 150 mM potassium aspartate solution, buffered to 1.1 M free Ca 2ϩ , and the holding potential was Ϫ95 mV after correcting for junction potentials (see "Experimental Procedures"). Upper panel, total current obtained by applying 300-ms steps at 20-mV intervals between Ϫ65 and ϩ15 mV. Control current was measured 5 min after attaining the whole-cell configuration, and then margatoxin (MgTx, 2.5 nM) was added to the bath at 15 min to block Kv1.3, and charybdotoxin (ChTx, 15 nM) was added at 30 min to block K Ca , leaving the "leak" (anion) current. Lower panel, current versus voltage (I-V) relations obtained from the same cell using a voltage ramp protocol, before and after addition of the peptide toxins, as above. All voltage ramps began with a step to Ϫ125 mV, followed immediately by a 225-or 300-ms ramp to ϩ40 mV. and hIK1/hKCa4 channels expressed in Xenopus oocytes (K d ϳ25 nM (8)) and HEK cells (K d 387 nM (9)). Clotrimazole effectively blocked IK in lymphoblasts (IC 50 ϭ 40 nM) and hSK4 in CHO cells (IC 50 ϭ 56 nM), with no time or voltage dependence (Fig. 2, C and D). Although some brain SK channels are potently blocked by apamin and d-tubocurarine (26,27), neither hSK4 nor the lymphoblast IK current was sensitive to these drugs (apamin, IC 50 Ͼ Ͼ100 nM; d-tubocurarine, IC 50 Ͼ Ͼ250 M). Iberiotoxin, a potent inhibitor of large conductance K Ca channels, did not block hSK4 or IK (IC 50 Ͼ200 nM, data not shown).
For functional studies of proliferation and volume regulation in lymphocytes, we avoided ChTx since it blocks both IK and Kv1.3 currents with similar potencies. Margatoxin is much more selective for Kv1.3 but can reduce IK at high concentrations (Ͼ10 nM, data not shown); thus we used agitoxin-2, which is more potent (K d ϳ200 pM) and does not block IK (28). IK was selectively blocked with clotrimazole. Drug concentrations were chosen to block different amounts of the two currents. To eliminate essentially all Kv1.3, we used 5 nM AgTx-2 (ϳ25 K d ), whereas to allow us to test for additive effects of IK block, we used 250 nM clotrimazole (ϳ6 K d for IK). Peptide toxins are especially useful since they are not membrane-permeant, so are unlikely to affect other intracellular processes. Although clotrimazole is membrane-permeant and also inhibits cytochrome P-450 (25), there is no evidence for P-450 involvement in lymphocyte proliferation or volume regulation.
Proliferation Is Inhibited by Combined IK and Kv1.3 Channel Block-We first stimulated freshly isolated (naive) human T cells with the mitogen, phytohemagglutinin (PHA-P), and then measured proliferation after 3 days. Aliquots of these stimulated cells (lymphoblasts) were re-exposed to the mitogen for a further 3 days. In control experiments, 3 days after mitogen treatment the total nucleic acid content increased by 3.1and 3.3-fold for naive and re-stimulated cells, respectively. Neither AgTx-2 nor clotrimazole alone significantly inhibited proliferation of naive cells, whereas the combined drugs were effective, inhibiting by 36.5% (Fig. 3). Consistent with our prediction, proliferation of activated lymphoblasts was inhibited to a much greater degree by blocking IK. That is clotrimazole (250 nM, ϳ6 K d ) inhibited by 65.0%, whereas AgTx-2 (5 nM, ϳ25 K d ) inhibited by 18.4%. The two drugs were essentially additive, inhibiting by 86.8%. In parallel experiments using trypan blue exclusion, we determined that cell viability at the MgTx was present to block Kv1.3 current. A, hSK4 whole-cell I-V relations were plotted from currents in response to voltage steps, whereas native IK currents were from voltage ramps (as in Fig. 1D). The bath K ϩ concentrations were 5 or 145 mM for CHO cells, 5 or 150 mM for lymphoblasts, yielding calculated E K values of Ϫ86 mV and ϳ0 mV. B, charybdotoxin blocks hSK4 and IK currents. Currents in response to voltage clamp steps between Ϫ120 and ϩ20 mV show profound block by 20 nM ChTx, with a slight relief over time at positive potentials in CHO cells. C, clotrimazole blocks both hSK4 and IK in a voltage-and time-independent manner. Currents in response to voltage clamp steps as in B. D, dose-response curves show fraction of current remaining (mean Ϯ S.E.).

FIG. 3. Comparison of the ability of K ؉ channel blockers to inhibit proliferation of naive T cells (A) and previously activated T lymphoblasts (B).
The term, naive, is used to indicate cells that were stimulated from an initial resting state. Each well of a 96-well plate was seeded with 2 ϫ 10 4 resting cells or lymphoblasts and incubated with or without channel blockers for 10 min (5 nM agitoxin-2, AgTx-2; 250 nM clotrimazole, CLT). Then PHA-P (7 g/ml) was added to initiate or restimulate proliferation. After 72 h, the CyQUANT assay was used to measure a change in fluorescence that is proportional to the change in cell number (see "Experimental Procedures"). Data are expressed as mean Ϯ S.D. of four independent experiments (four replicates each). A Bonferroni multiple comparison test was used to assess each combination of treatments. Values that differ significantly from controls are indicated (***, p Ͻ 0.001), as are significant differences between drug treatments ( † † †, p Ͻ 0.001). end of the 3-day proliferation period was not reduced by the highest concentrations of channel blockers used but remained at Ͼ98%. Thus, reduced nucleic acid content reflects reduced proliferation and not cell death.
Regulatory Volume Decrease (RVD) Is Inhibited by K ϩ Channel Blockers-When T lymphocytes are swollen by exposure to hypotonic solution, they undergo a regulatory volume decrease (RVD) by loss of K ϩ and Cl Ϫ through ion channels, followed by osmotically obliged water loss (see "Discussion"). To assess the relative contributions of IK and Kv1.3 to RVD, we exposed cells to a standard hypotonic shock (56% of normal osmolarity) in the presence or absence of 5 nM AgTx-2, 250 nM clotrimazole, or both drugs. There were no changes in cell volume (with or without blockers) measured for up to 15 min in isotonic solutions. Following hypotonic shock, maximal swelling for each blocker was expressed as a percent increase over the control volume in the same blocker. Both resting and activated T cells swelled to a maximal volume within 2 min and then recovered to varying degrees depending on which K ϩ channels were blocked. Blocking IK (clotrimazole), Kv1.3 (AgTx-2), or both channels increased the maximal swelling of resting T cells (Fig.  4A). For lymphoblasts, blocking IK was as effective as blocking both channels, whereas blocking Kv1.3 alone did not significantly increase the maximal volume. Hence, in lymphoblasts, IK appears to play a greater role than Kv1.3 during the initial swelling phase.
We determined the effects of channel blockers on RVD within the first 6 min after the hypotonic shock (Fig. 4B). By this time, in control hypotonic solution both cell types had almost fully recovered from swelling: by 81.1 Ϯ 6.8% (n ϭ 6) in resting cells and 88.5 Ϯ 5.5% (n ϭ 6) in lymphoblasts. K ϩ channel blockers attenuated this recovery. Blocking Kv1.3 reduced volume recovery to 62.3 Ϯ 9.0% (n ϭ 4) in resting cells and 80.9 Ϯ 5.9% (n ϭ 4) in activated lymphoblasts, a significantly greater effect in resting cells. IK block was more effective than Kv1.3 block for both cell types: recovery was 39.6 Ϯ 6.5% (n ϭ 5) in resting cells and 28.8 Ϯ 5.3% (n ϭ 6) in lymphoblasts, a significantly greater effect in lymphoblasts. When both channels were blocked, RVD was further decreased in resting cells (30.7 Ϯ 6.1% recovery, n ϭ 5) but not in lymphoblasts (29.8 Ϯ 6.8% recovery, n ϭ 5). Our data show that both K ϩ channels contribute to RVD; IK is generally more important and, as predicted from its up-regulated expression, IK plays a greater role than Kv1.3 in lymphoblasts.
Calmodulin Antagonists Inhibit Native IK and Expressed hSK4 Currents-Three structurally unrelated drugs were used to inhibit the Ca 2ϩ -calmodulin complex: trifluoperazine (TFP, K d ϳ1 M (29)), W-7 (K d ϳ12 M (30)), and calmidazolium (K d ϳ50 nM (31)). None of the drugs were toxic at the concentrations used; i.e. the cells did not become leaky even after a 30 -60-min preincubation. Representative current traces are shown in Fig. 5, A-C, and the voltage dependence is summarized in Fig. 5D by plotting the current amplitude at the end of each step as a function of membrane potential. For hSK4 current in stably transfected CHO cells, TFP and W-7 effects were steeply voltage-dependent, with increased current at negative potentials (36% for TFP and 11% for W-7 at Ϫ100 mV). The inhibition at positive potentials (75% for TFP and 71% for W-7 at ϩ40 mV) showed a pronounced time dependence (Fig. 5, A  and B). Calmidazolium was not voltage-dependent; i.e. 500 nM inhibited hSK4 by ϳ35% at all voltages. For native IK current we were particularly interested in drug effects at negative membrane potentials typical of a non-excitable lymphocyte. In contrast to hSK4, the lymphoblast IK current was significantly decreased at negative potentials by all three calmodulin antagonists. IK inhibition by 10 M TFP was mildly voltage-dependent (a 55% decrease at Ϫ100 mV versus 75% at ϩ40 mV), and this was due to a time-dependent reduction at depolarized potentials (Fig. 5A). Effects of 25 M W-7 were qualitatively similar, with 42% inhibition at Ϫ100 mV and 65% at ϩ40 mV. Again, inhibition by calmidazolium was not voltage-dependent and 500 nM inhibited IK by ϳ70% (Fig. 5, C and D). Thus, calmidazolium more effectively inhibited native IK than hSK4, and at negative potentials TFP and W-7 were only effective in reducing the native IK. This suggests two mechanisms of action, a physiologically relevant reduction of native IK at negative potentials and a time-and voltage-dependent reduction of both IK and expressed hSK4 at positive potentials.
To investigate further IK inhibition by CaM antagonists in lymphoblasts, we used voltage ramps or steps (see Figs. 1D and 2) and measured IK at a sufficiently negative potential to elicit large IK currents without contamination by Kv1.3. The current amplitude was measured at Ϫ120 mV from each cell before (control) and after adding a calmodulin antagonist to the bath. At the end of each experiment the anion current was recorded (after blocking IK and Kv1.3 with 20 nM ChTx) and subtracted from each total current at Ϫ120 mV to calculate the IK amplitude. Both ramp and step protocols yielded the same results. As summarized in Fig. 6, all three CaM antagonists dose-dependently inhibited IK (n ϭ 4 -6 cells unless otherwise indicated). When bath-applied during a recording, TFP reduced IK by

FIG. 4. Inhibition of the RVD by K ؉ channel blockers in resting T cells compared with activated lymphoblasts.
Flow cytometric analysis of right angle light scattering was used to measure cell swelling (see "Experimental Procedures"). A, maximal swelling, which occurred within 2 min after the hypotonic shock, is expressed as the percent increase above the control value. Control values for each drug treatment were measured in isotonic solution containing the drug, that is 5 nM AgTx-2 to block Kv1.3, 250 nM clotrimazole to block IK, or both drugs. Open bars, resting T cells; hatched bars, activated T lymphoblasts 3 days after mitogenic stimulation. B, the percent recovery at 6 min after hypotonic shock. Data are presented as mean Ϯ S.D. from at least 4 experiments. Bonferroni multiple comparison tests were used to assess all differences. Significant differences between control and drugtreated cells are indicated for resting T cells (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001) or lymphoblasts ( †, p Ͻ 0.05; † †, p Ͻ 0.01; † † †, p Ͻ 0.001). We used a competition experiment in which excess CaM (50 M with 1.1 M free Ca 2ϩ , see "Experimental Procedures") was added to the pipette solution, followed by 10 M W-7 addition to the bath, in the expectation that CaM would bind to internal W-7 and relieve inhibition. CaM addition did not significantly increase IK compared with the current in the same cells during the first 2 min of recording; however, the variability between cells was high (IK increased by 47 Ϯ 26%, p Ͼ 0.34, n ϭ 6). Nevertheless, as expected, excess CaM reduced the W-7-induced IK inhibition (from 84.2 Ϯ 5.2% to 11.3 Ϯ 29.1%, n ϭ 12, p Ͻ 0.05), consistent with CaM-antagonist competition at an intracellular site. This result also rules out drug effects at external sites on the channel at negative potentials (see "Discussion"). When calmidazolium was bath-applied after a recording was begun, it dose-dependently inhibited native IK current within 5-10 min, by 54.2 Ϯ 4.5% at 100 nM (p Ͻ 0.05) and by 69.0 Ϯ 3.1% (p Ͻ 0.01) at 500 nM. Inhibition was more effective when lymphoblasts were preincubated for 10 -15 min before recording (94.1 Ϯ 5.8% inhibition, p Ͻ 0.01). In contrast, a 1-2-h preincubation with 500 nM calmidazolium did increase the inhibition of hSK4 in CHO cells (34%, n ϭ 8). Thus, slow drug permeation, which is a potential limiting factor, does not explain the difference in sensitivities between native IK and expressed hSK4 currents. Calmodulin Binds to a Proximal Portion of the C Terminus of hSK4 -One possibility is that CaM antagonists affect this K Ca current by interfering with interactions between CaM and the channel protein. If so, by analogy with properties of CaM binding to CaMK, one might expect CaM-hSK4 channel binding to be Ca 2ϩ -dependent and competitively inhibited by CaM antagonists. We tagged hSK4 with the Flag epitope and expressed this construct stably in CHO cells and then determined whether the expressed protein bound to CaM-conjugated agarose beads. This assay favors detection of proteins that reversibly bind to CaM. That is if hSK4 channels had already bound to CaM in situ (as reported for brain SK channels (15)), they would be unable to interact with CaM on the agarose beads. The membrane proteins were then incubated with CaM-agarose beads in the presence or absence of Ca 2ϩ or the CaM antagonist, calmidazolium.

hSK4/hIK1 Mediates Proliferation and RVD in T Cells
After pre-clearing the samples and washing the beads to remove nonspecific binding, Ponceau-stained Western blots indicated that several proteins bound to the beads (data not shown). However, in samples containing the expressed fulllength Flag-tagged hSK4 protein, only a single band was labeled by the anti-Flag antibody. Fig. 7A shows that binding of the wild-type full-length hSK4 protein was Ca 2ϩ -dependent (i.e. greatly reduced by EGTA) but not competed by 1,000 nM of the CaM antagonist, calmidazolium. To map the region of the

FIG. 5. Calmodulin antagonists reduce hSK4 stably expressed in CHO cells and native IK current in activated lymphoblasts.
A-C, representative currents in control saline or with 10 M trifluoperazine (TFP), 25 M W-7, or 500 nM calmidazolium added to the bath. D, average current in CHO cells (q) and lymphoblasts (Ⅺ) expressed as current at the end of each voltage clamp step as a fraction of control current (I Drug /I control ) at each membrane potential (mean Ϯ S.E., 2-6 cells). Lymphoblast IK currents were omitted near the reversal potential (about Ϫ86 mV) when they were too small to construct accurate ratios. Some error bars are smaller than the symbol. All values marked * are significantly lower for IK than respective values for hSK4 currents (p Ͻ 0.05).
FIG. 6. Inhibition of lymphoblast IK by calmodulin antagonists. IK amplitude at Ϫ120 mV was calculated from whole-cell currents in response to voltage ramps as explained in the text. The bars show mean Ϯ S.E. for 4 -6 cells unless otherwise indicated under "Results." A, after the onset of each recording, trifluoperazine was added to the bath in increasing concentrations (1, 2.5, and 5 M) at 10-min intervals. A separate set of cells was used to control for current rundown by adding 10 M trifluoperazine to the bath immediately after the recording had stabilized. Each remaining steady-state current was calculated as a percent of the control value for that cell (* indicates a significant reduction of IK, see text). B, increasing doses of W-7 were added to the bath with each cell serving as its own control (*, significant reduction of IK). In a separate series of cells, 50 M CaM was present in the pipette, and then 10 M W-7 was applied to the bath 15-20 min after a whole-cell recording was established. Intracellular CaM significantly abrogated the effect of W-7 ( †). C, after measuring control currents, increasing concentrations of calmidazolium were added to the bath at 10-min intervals (2nd to 4th bars). In separate experiments cells were preincubated with 500 nM calmidazolium for 10 -15 min before recording for 10 -15 min in the continued presence of the drug. † indicates significantly greater inhibition following preincubation.

hSK4/hIK1 Mediates Proliferation and RVD in T Cells
channel that binds to CaM, we made several constructs by deleting the following: the distal C terminus (leucine zipper region, which we call "N-M-Ct1" since it contains the N terminus, membrane-spanning, and C terminus 1 regions), the entire N terminus and transmembrane domains (Ct1-Ct2), all but the proximal C terminus (Ct1), or the proximal C-terminal tail (N-M-Ct2). All but one channel construct bound to CaM in a Ca 2ϩ -dependent manner that was not competitive with calmidazolium, i.e. the construct lacking the proximal C terminus (Ct1) did not detectably bind to CaM.
C-terminal Deletions Have Functional Consequences-We used the membrane potential (V m ) to monitor the expression of functional hSK4 channels of various constructs in transfected CHO cells (Fig. 7B). Whole-cell patch clamp recordings were made with 1 M free Ca 2ϩ in the pipette (as for all recordings of currents), and the membrane potential was recorded in the current clamp mode. The V m of CHO cells transfected with wild-type hSK4 channels was Ϫ67 Ϯ 4 mV (n ϭ 5) indicating a significant K ϩ permeability (Nernst potential for K ϩ , Ϫ82 mV), and hSK4 with a C-terminal Flag tag produced the same V m (Ϫ68 Ϯ 4 mV, n ϭ 5). The V m of control cells transfected with vector alone (pcDNA3) was Ϫ10 Ϯ 8 mV, indicating a very low background K ϩ permeability. The hSK4 construct lacking the CaM-binding region (N-M-Ct2) did not produce a negative membrane potential despite the high intracellular Ca 2ϩ ; V m was significantly less negative than for the wild-type construct (Ϫ2 Ϯ 1 mV, n ϭ 5, p Ͻ 0.001). Interestingly, V m was also close to 0 (Ϫ5 Ϯ 2 mV, n ϭ 6) for the construct lacking the distal C terminus (N-M-Ct1).
The Lymphoblast IK Current Is Reduced by a CaM Kinase Antagonist-KN-62 is a membrane-permeant CaM kinase antagonist, whereas KN-04 is an inactive analogue used as a negative control. KN-62 reduced IK in a time-and voltageindependent manner (Fig. 8A). Fig. 8B summarizes effects of KN-62 alone or in combination with the CaM antagonist, W-7. IK was measured in lymphoblasts at Ϫ120 mV after subtracting the small anion current (as for Fig. 6). We first tested intact cells at 37°C in an attempt to maintain kinase and phosphatase activity and turnover of protein phosphorylation. Preincubation with 10 M KN-62 alone (30 min, 37°C) decreased the current by 55.5 Ϯ 11.2% (p Ͻ 0.02, n ϭ 4), whereas KN-04 had no effect (97 Ϯ 11% of the control value, n ϭ 4, p Ͼ 0.8). After KN-62 preincubation and with the drug present throughout whole-cell recordings, adding W-7 to the bath further reduced the current, i.e. by 80.2 Ϯ 10.1% reduction (n ϭ 4, p Ͻ 0.05) with the combined drugs. Acute effects of bath-applied KN-62 were tested during whole-cell recordings at ϳ37°C. IK was substantially reduced by KN-62 (by 67.9 Ϯ 11.7%; n ϭ 4, p Ͻ 0.02) but not by KN-04. Under drug-free conditions, IK in lymphoblasts was not obviously temperature-dependent; the specific conductance was 606 Ϯ 210 pS/picofarads at room temperature and 552 Ϯ 13 pS/picofarads when the same cells were warmed to Ͼ33°C (n ϭ 6, p Ͼ 0.7). Interestingly, the variability in current amplitude was greatly reduced at the higher temperature. Unlike the native channels in lymphoblasts, hSK4 stably expressed in CHO cells was not inhibited by 10 M KN-62 and remained at 94.3 Ϯ 2.5% (n ϭ 6, p Ͼ 0.2) of the control value measured with solvent alone (1% Me 2 SO).

Comparison of the Cloned hSK4 with IK Current in Activated T Cells
The present results are entirely consistent with the IK current in T lymphocytes being the product of the hSK4/hIK1/ hKCa4 gene, which was recently cloned from cDNA libraries from human placenta (7), pancreas (8), and lymph node (9). Since the product we cloned is 100% identical to hSK4/hIK1/ hKCa4, differences in properties of the native lymphocyte IK and exogenously expressed hSK4 channels are not expected unless such properties are determined by something other than the ␣ subunit of the channel. In principle, differences could arise if the channel forms heteromultimers with another protein, if alternative splice variants exist, or if the channel interacts with accessory molecules. It is intriguing that multiple transcript sizes are commonly seen for this channel, i.e. 2.6 and 3.8 kb (7), ϳ2.1 kb, and at least one larger band (8), 2.2 kb, with two larger bands (9), and a prominent 2.2-kb band with a weaker 2.6-kb band (present study, data not shown).
Biophysical properties of the lymphocyte IK current have been described at the single channel (2,5,24) and whole-cell level (5). Whereas channel gating is independent of voltage, it is highly sensitive to intracellular free Ca 2ϩ , activating at Ͻ200 nM in T and B cells, reaching half-maximal activation at about 450 nM, and maximal activation at ϳ1 M (2, 5). Both the single channel and whole-cell current versus voltage (I-V) relations are inwardly rectifying with symmetrical K ϩ concentrations on both sides of the membrane (5,6,24). The single channel I-V relation is linear under physiological Na ϩ /K ϩ gradients, which, together with the voltage-independent gating, results in a whole-cell current that is nearly linear (5,24). Expressed hSK4/hIK1/hKCa4 currents (Refs. 7-9 and present study) have the following features in common with the lymphocyte IK current (Refs. 2, 3, 5, 7, and 24 and present study); activation by sub-micromolar free Ca 2ϩ , time-and voltageindependent gating, inwardly rectified single channel I-V relations in symmetrical K ϩ (10 -35 pS), and nearly linear I-Vs in physiological Na ϩ /K ϩ gradients (ϳ10 pS).
The pharmacological profiles of native IK and hSK4/hIK1/ hKCa4 are also similar. Native IK in lymphoblasts and hSK4 expressed in CHO cells were blocked by ChTx (IC 50 2-10 nM) but very poorly by iberiotoxin (IC 50 Ͼ200 nM), margatoxin (IC 50 Ͼ100 nM), or tetraethylammonium (IC 50 30 -40 mM). Clotrimazole showed a similar potency for blocking IK in lymphoblasts (present study) and for heterologously expressed hSK4 (IC 50 25-60 nM (Ref. 8 and present study)). The hSK4/hIK1/hKCa4 channel is expected to be insensitive to both apamin and dtubocurarine since it lacks two necessary amino acids in the putative pore (27) and, as expected, neither the lymphoblast IK nor the expressed hSK4 current were significantly inhibited by apamin (IC 50 Ͼ100 nM (Refs. 5 and 7-9 and present study) or d-tubocurarine (IC 50 Ͼ250 M; present study).

Increased Role for hSK4 in Lymphoblast Proliferation
K ϩ channel activity is important during the early activation phase of naive T cells, especially for maintaining a hyperpolarized membrane potential, promoting a rise in intracellular Ca 2ϩ , and permitting a cascade of events that culminates in interleukin-2 production (1,32,33). In the first few hours after mitogenic stimulation, precisely when Ca 2ϩ elevation is necessary (34,35), K ϩ channel blockers, or other means of depolarizing T cells (high external K ϩ , voltage clamp), inhibit T cell activation (36) by compromising Ca 2ϩ influx and the resulting rise in Ca 2ϩ . Early studies using non-selective K ϩ channel blockers (e.g. quinidine, 4-aminopyridine) were later substantiated by more selective peptide toxins including charybdotoxin, which blocks both IK (K d ϳ2-6 nM (Refs. 3 and 5 and present study)) and Kv1.3 channels (K d ϳ1 nM (1,5,16,17,24)), and margatoxin or noxiustoxin which block Kv1.3 but not IK channels (1,16,17,28,37). From the limited functional studies using blockers that discriminate between Kv1.3 and other K ϩ channels, Kv1.3 appears to be important for activation of naive T cells through pathways that are Ca 2ϩ -dependent (1,32).
The contribution of K Ca channels to T cell activation and proliferation is still poorly understood. Since we previously found that mitogens activate K Ca channels in cell-attached patches from naive human T cells, we proposed that they also play a role in T cell activation (14). In the present study we observed a 14.6-fold increase in hSK4 transcripts by 3-4 days after mitogenic stimulation. During the same period, mRNA levels for Kv1.3 increased only 1.3-fold. These changes are consistent with previous patch clamp studies showing an approximate doubling in Kv1.3 current (see Ref. 1), a 30-fold increase in the ChTx-sensitive K Ca current (5) and an increase in hKCa4 mRNA (9). The prediction that IK will be increasingly important for the secondary immune response (e.g. proliferation of previously activated lymphoblasts) is supported by the present results. There is also a recent report that Ca 2ϩ signaling and proliferation were more strongly inhibited by ChTx in lymphoblasts than in naive T cells (38); however, ChTx does not discriminate between IK and Kv1.3 channels. To separate better the contributions of Kv1.3 and IK to T cell function, we used AgTx-2 to block Kv1.3 and clotrimazole to block IK. Consistent with our expectations, IK block more effectively inhibited proliferation of lymphoblasts than naive T cells. Furthermore, despite the greater Kv1.3 channel block (AgTx-2 at ϳ25 K d ) than IK block (clotrimazole at ϳ6 K d ), IK block was more effective in inhibiting lymphoblast proliferation, i.e. by 65.0% compared with 18.4% for Kv1.3 block. Blocking both channels (AgTx-2 ϩ clotrimazole) was approximately additive, reducing lymphoblast proliferation by 86.8%. Proliferation of naive T cells was also sensitive to blocking both channels (36.5% inhibition), consistent with previous reports of reduced proliferation when IK ϩ Kv1.3 were blocked with ChTx (1,32,33).
How might both Kv1.3 and IK channels contribute to T cell proliferation? Within seconds after stimulating the T cell receptor, tyrosine-kinase mediated activation of phospholipase C produces inositol 1,4,5-trisphosphate and quickly triggers Ca 2ϩ release from internal stores. A plasma membrane channel (the Ca 2ϩ release-activated Ca 2ϩ channel) then opens to allow Ca 2ϩ influx, which is required for several hours. Ca 2ϩ release-activated Ca 2ϩ channel opening is not voltage-dependent, but Ca 2ϩ influx is strongly driven by the membrane potential. Thus, any means of increasing the K ϩ conductance and hyperpolarizing the cell will facilitate Ca 2ϩ entry. Kv1.3 is voltage-gated, activated by depolarization, and its steady-state activity is maximal between Ϫ50 and Ϫ30 mV in resting T cells, depending on post-translational modulation (37). Hence it is likely to play a role only when the membrane is moderately depolarized. In contrast, gating of the IK/hSK4 channel is voltage-independent but exquisitely sensitive to internal Ca 2ϩ ; thus, it is well designed to open whenever Ca 2ϩ rises marginally above the resting level. Rather than Ca 2ϩ changing in a sustained manner after T cell receptor stimulation, Ca 2ϩ and membrane potential can oscillate (1,38,39). Collectively, these complementary properties would allow the cell alternately to use IK channels when Ca 2ϩ is high, even if the membrane is hyperpolarized, and Kv1.3 channels during periods of low Ca 2ϩ and/or depolarization.

Role of hSK4 and Kv1.3 in Volume Regulation
Volume regulation in leukocytes and other mammalian cells has been extensively reviewed (10 -12). The RVD involves ion efflux through separate K ϩ and anion channels. Its rate and extent depend on the combined ion conductances; thus if either the K ϩ or Cl Ϫ current is small (or blocked pharmacologically) it will limit volume recovery. Identifying the particular K ϩ channel(s) that underlie RVD has been problematic, largely due to the lack of selective blockers and uncertainty over whether swelling evokes a rise in intracellular Ca 2ϩ . For instance, early evidence of a role for Kv1.3 was not convincing since it first relied on nonspecific K ϩ channel blockers and then on ChTx (10,11,40) which we now know blocks both Kv1.3 and IK. Kv1.3 can confer RVD when transfected into a mouse T cell line that lacks voltage-gated K ϩ currents (13); however, that study did not address the presence or role of K Ca channels. To determine whether intracellular Ca 2ϩ is elevated in human T cells during RVD, we previously designed a perfused cuvette system for fluorometric measurements (41). We found that hypotonic shock elicits a rapid, biphasic rise in Ca 2ϩ (14) which comprises release from internal stores and influx across the plasma membrane. Ca 2ϩ reached a peak in 1-2 min and remained elevated for at least 10 min, and the entire pattern was indistinguishable from Ca 2ϩ signaling during T cell activation. Thus, we proposed (14) that both K Ca and Kv1.3 channels will contribute to RVD, with Ca 2ϩ and voltage oscillations alternately opening each type of K ϩ channel.
Previous studies of RVD in T cells have been restricted to resting cells and show a stereotypical response (10,11,13); within 1-2 min after a hypotonic shock, T cells swelled to ϳ120% of their original volume and then returned to their original volume within 5-15 min. Our results, using right angle light scattering to measure RVD, are in excellent agreement both in the extent of swelling (ϳ120%) and in the time course, with maximal swelling within 1-2 min and nearly full recovery within 6 min for both resting T cells and lymphoblasts. We have now assessed the relative contributions of IK and Kv1.3 channels to RVD in resting T cells compared with activated lymphoblasts. Owing to the dramatic increase in IK and small increase in Kv1.3 in lymphoblasts, we expected RVD to depend more on IK than Kv1.3 channels in lymphoblasts.
We examined the role of each channel type in the initial swelling to maximal volume and in the degree of recovery by 6 min after a 56% hypotonic shock. For resting T cells, blocking Kv1.3 or IK, or both channels simultaneously, increased the maximal volume, implying that K ϩ efflux through both channels occurs even during the initial swelling phase. Consistent with this conclusion, in previous studies T cells swelled less than predicted for a passive osmometer (10 -12). We had anticipated a role for IK channels in resting cells, since IK is expressed (Refs. 6 and 39 and present study), and intracellular Ca 2ϩ rises after a hypotonic shock (14). For lymphoblasts, IK block was very effective in increasing the maximal volume, whereas Kv1.3 block had no effect. Thus, in lymphoblasts volume regulation also proceeds during the swelling phase but IK plays a much greater role than Kv1.3 current at this time.
The extent of recovery after the maximal volume is reached reflects both the K ϩ and Cl Ϫ conductances during the RVD phase. Substantial recovery occurred within 6 min in both cell types, and the slightly greater recovery in lymphoblasts is consistent with up-regulation of IK/hSK4 expression. RVD was significantly inhibited by blocking Kv1.3 channels in both resting T cells and lymphoblasts but was more effective in resting cells. For both cell types IK block was more effective than Kv1.3 block; moreover, inhibition of RVD was greater in lymphoblasts. Thus, both K ϩ channels contribute to RVD, but their relative importance is opposite as follows: Kv1.3 plays a greater role in resting cells, and IK is more important in lymphoblasts. Not only are these results predicted from the up-regulated expression of IK/hSK4 in lymphoblasts but IK activation implies that hypotonic shock elicits an early and sustained rise in Ca 2ϩ in activated lymphoblasts, as we have previously shown for resting T cells (14).
Although there is insufficient information to calculate K ϩ fluxes through Kv1.3 and IK channels during RVD, some predictions can be made by considering their expression and biophysical properties. Flux through each channel type is proportional to the number of channels (n), their open probability (P o ), their single channel conductance (␥), and the driving force, which is the same at a given voltage. For resting T cells, the number of Kv1.3 channels per cell is much larger than IK/hSK4 channels, and ␥ is similar (ϳ10 pS in a normal Na/K gradient). So, for IK channels to play a substantial role in RVD, their P o must be much larger than the P o of Kv1.3 channels in resting cells. Kv1.3 contribution will be controlled by the membrane potential since these channels require moderate depolarization to be tonically active (37,42,43). A simple model is that Ca 2ϩ remains elevated thereby activating IK, and the membrane potential remains hyperpolarized, thereby reducing the opening of Kv1.3 channels. IK/hSK4 expression is much higher in lymphoblasts than in resting T cells, and since IK gating is voltage-independent, this channel is expected to contribute more whether or not the membrane potential fluctuates, provided Ca 2ϩ remains modestly elevated.

Calmodulin-dependent Modulation, Evidence for More Than One Mechanism
Our electrophysiological results implicate calmodulin (CaM) in regulating native IK channels in lymphoblasts. In principle, CaM antagonists could act by interfering with interactions between CaM and the channel protein, by interacting with accessory CaM-binding molecules (e.g. CaM kinases or channel ␤ subunits, if they exist), or by directly interfering with the hSK4/hIK1 Mediates Proliferation and RVD in T Cells channel protein. Some predictions can be made from the mechanism by which the antagonists inhibit CaM (44 -46). In cellfree systems, CaM changes conformation when at least two of its four Ca 2ϩ -binding domains are saturated (K d ϳ2.4 M Ca 2ϩ ). CaM antagonists can then bind reversibly to a newly exposed hydrophobic site (47) thereby preventing interactions between CaM and target proteins. Thus, it is expected that excess CaM will competitively reduce inhibition by titrating the amount of drug available for inhibition.
Direct interactions of some CaM antagonists in the pore of some K ϩ channels have been proposed when their potency for CaM inhibition differed from that of channel inhibition (48 -50), or the drugs were effective even when Ca 2ϩ was not elevated (50), or exogenously added CaM did not compete with the antagonists (49). We found that trifluoperazine and W-7 produced time-and voltage-dependent decreases in both native IK and expressed hSK4 current at positive potentials, which may reflect a direct drug interaction with the channel protein. Of greater physiological relevance is the inhibition of native IK channels we observed at negative membrane potentials. In this case the potency of inhibition by W-7 and TFP was consistent with effects on CaM, and as expected for competitive drug binding, excess internal CaM significantly relieved the inhibition by W-7. This result also rules out significant channel block by W-7 from the outside at negative potentials. CaM antagonists may affect the lymphoblast IK current through interactions between CaM or other CaM-binding molecules and the channel protein. Such interactions must differ for hSK4 channels stably expressed in CHO cells since, at negative potentials, these currents were not inhibited by TFP or W-7, and calmidazolium was less effective than on IK currents. As discussed below, differences in actions on native IK and hSK4 channels may reflect multiple sites of action.
Direct Interactions between CaM and IK/hSK4 Channels-Despite the exquisite Ca 2ϩ sensitivity of IK/hSK4 gating, the primary amino acid sequence of the ␣ subunit contains no known Ca 2ϩ -binding sites, that is no E-F hands, C2 domains (51), or Ca 2ϩ "bowls" (52). We found that channels made from wild-type, full-length hSK4 ␣ subunits bind to calmodulin. Although binding was greatly inhibited when Ca 2ϩ was chelated, some binding remained. Deletion mutants of several cytoplasmic regions that are relatively conserved between hSK4 and brain SK channels showed that CaM binding was restricted to the proximal C-terminal tail of hSK4 (a region we call "Ct1," see Fig. 7). Deleting the Ct1 region prevented the expression of functional Ca 2ϩ -gated hSK4 channels. When wild-type hSK4 was expressed and whole-cell membrane potential (V m ) recordings were made with micromolar intracellular Ca 2ϩ to maximally activate hSK4, V m became highly negative owing to the hyperpolarizing K ϩ conductance. In contrast, the Ct1-deleted channel failed to produce a hyperpolarizing K ϩ conductance, and V m remained essentially at zero. There are two possibilities as follows: without CaM binding the channels did not open in response to high Ca 2ϩ (as is the case for SK2 channels (15)) or the mutant channels did not assemble properly in the cell membrane. In the future it will be useful to examine the assembly and trafficking of mutant hSK4 channels, particularly since the Ct2-deletion mutant (lacking the leucine zipper region) also failed to produce a hyperpolarizing K ϩ conductance. The ␣ subunits of brain SK channels (SK1-3) also bind to CaM in the proximal part of the cytoplasmic C terminus, and CaM apparently serves as the Ca 2ϩ -binding gate (15). Most of the C terminus of SK2 channels (4 ␣ helices, A-D) bound to CaM, whereas a "post-D" C-terminal tail (corresponding to our Ct2, leucine-zipper region) did not. If helices A-D were all present, binding was independent of Ca 2ϩ , whereas helices B and C and B-D conferred Ca 2ϩ -dependent binding to CaM.
Our results on hSK4 share several features with SK2 (15), but they also differ in ways that are consistent with additional sites of interaction or modulation of the native channel in lymphocytes. Although brain SK channels have only modest overall homology (ϳ40%) to hSK4 (7), some regions are more homologous. The 95-amino acid CaM-binding domain (Ct1) that we identified in hSK4 is the same channel region as helices A-C in SK2. Helix A in SK2 is highly homologous to the corresponding region of hSK4 (79% identical), whereas regions B and C have much lower homology (20% identical). For SK2, the Ca 2ϩ -independent CaM binding and patch clamp studies in which calmidazolium failed to inhibit the expressed channels were taken as evidence that CaM binds constitutively and irreversibly to brain SK channels (15). We found that most of the CaM binding to hSK4 was Ca 2ϩ -dependent; however, as explained earlier, the binding assay we used would not detect channels irreversibly bound to CaM. Several possible explanations for these differences will require further study. For instance, there may be more than one CaM-binding site with different affinities, as is the case for cation channels in retinal rods (53), with a lower affinity site that is reversible and Ca 2ϩ -dependent, perhaps in Ct1 of hSK4 (helices B-D in SK2). CaM antagonists inhibited the native current in lymphoblasts, with little or no inhibition of expressed hSK4 currents, a result that is consistent with the failure of 1000 nM calmidazolium to prevent CaM binding to the expressed hSK4 protein. A further possibility is that weaker CaM channel binding in lymphoblasts, perhaps a result of other protein-channel interactions (see below), allows more effective competition by CaM antagonists.
Evidence for Accessory Molecules in Lymphocytes-The striking differences in inhibition at negative potentials of native IK versus expressed hSK4 channels provide the first evidence that accessory molecules (other than CaM) modulate a member of the SK channel family. Candidate molecules include CaMK, calcineurin, and ␤ subunits analogous to those interacting with voltage-gated K ϩ channels. Although ␤ subunits have not been identified for SK channels, there is evidence that apaminsensitive SK channels form hetero-oligomers. In a variety of cells expressing SK channels, apamin binds to both high (59 or 86 kDa) and low (30 or 33 kDa) molecular mass polypeptides that are integral membrane proteins (54). It is unlikely that such accessory molecules are essential for channel activity since all known members of the SK family, with the exception of rSK1 (7), are functional in expression systems, including Xenopus oocytes, HEK, and CHO cells. This observation also implies that the Ca 2ϩ -binding site, which is thought to be CaM bound to the channel (15), functions normally in these cells.
Since the CaM kinase antagonist, KN-62, inhibited native IK current but had no effect on hSK4, it is necessary to consider how CaM kinase might selectively modulate the current in lymphocytes. hSK4 contains a potential phosphorylation site in the C-terminal domain that should accommodate either CaM kinase II (or protein kinase A) (45); however, this site is not necessarily phosphorylated. It may be that CHO cells have insufficient CaM kinase (T cells express high levels of CaM kinase II and IV (55)) or that the site of phosphorylation is not on the channel protein itself, but rather on a ␤ subunit or other unknown accessory molecule. Interestingly, in lymphoblasts the increased inhibition by W-7 in the presence of KN-62 is consistent with dual modulation by CaM binding and CaM kinase.
Potential modulation of lymphocyte IK/hSK4 channels by CaM and CaM kinase is of broader importance. Early in T cell activation or lymphoblast re-activation, there is a rise in intracellular Ca 2ϩ that activates CaM-dependent enzymes. These include CaM kinases II and IV (55) and calcineurin (protein phosphatase 2B), which is highly expressed in lymphocytes and crucial for T cell proliferation (34). CaM antagonists can inhibit some lymphocyte functions that either trigger conductive K ϩ fluxes or are sensitive to the membrane potential of the cell (which depends on K ϩ channels). T cell activation (56), cellmediated cytotoxicity (57), and volume regulation (40) are inhibited both by K ϩ channel blockers and by CaM antagonists. Our present results provide new evidence that a specific K ϩ channel (IK/hSK4) that is important for at least two of these functions (proliferation and volume regulation) is susceptible to CaM antagonists, thus providing a link between CaM and K ϩ channels in regulating lymphocyte function.