Functionally diverse complement of large conductance calcium- and voltage-activated potassium channel (BK) alpha-subunits generated from a single site of splicing.

The pore-forming alpha-subunits of large conductance calcium- and voltage-activated potassium (BK) channels are encoded by a single gene that undergoes extensive alternative pre-mRNA splicing. However, the extent to which differential exon usage at a single site of splicing may confer functionally distinct properties on BK channels is largely unknown. Here we demonstrated that alternative splicing at site of splicing C2 in the mouse BK channel C terminus generates five distinct splice variants: ZERO, e20, e21(STREX), e22, and a novel variant deltae23. Splice variants display distinct patterns of tissue distribution with e21(STREX) expressed at the highest levels in adult endocrine tissues and e22 at embryonic stages of mouse development. deltae23 is not functionally expressed at the cell surface and acts as a dominant negative of cell surface expression by trapping other BK channel splice variant alpha-subunits in the endoplasmic reticulum and perinuclear compartments. Splice variants display a range of biophysical properties. e21(STREX) and e22 variants display a significant left shift (>20 mV at 1 microM [Ca2+]i) in half-maximal voltage of activation compared with ZERO and e20 as well as considerably slower rates of deactivation. Splice variants are differentially sensitive to phosphorylation by endogenous cAMP-dependent protein kinase; ZERO, e20, and e22 variants are all activated, whereas e21 (STREX) is the only variant that is inhibited. Thus alternative pre-mRNA splicing from a single site of splicing provides a mechanism to generate a physiologically diverse complement of BK channel alpha-subunits that differ dramatically in their tissue distribution, trafficking, and regulation.

Several sites of alternative pre-mRNA splicing within the mammalian BK channel ␣-subunit have been described, the majority of which are located within the intracellular C-terminal domain of the channel ␣-subunit (35). Analysis of individual alternatively spliced variants generated at distinct splice sites in different species has revealed that alternative pre-mRNA splicing can dramatically modify the functional properties of the BK channel ␣-subunit, including changes in calcium and voltage sensitivity (36 -42), regulation by protein phosphorylation (22,43,44), and other intracellular signaling cascades (45,46) as well as cell surface expression (47,48).
Although it is known that multiple alternatively spliced variants may be generated from a single site of splicing, the extent of the phenotypic diversity that can be generated by these splicing decisions is largely unknown. In this study we have performed transcript scanning and functional expression to isolate and analyze the properties of five splice variants generated by alternative pre-mRNA splicing at a single site of splicing, the mammalian site C2. This site of splicing is located in the intracellular C terminus of the ␣-subunit within the linker region between the two putative regulators of potassium conductance domains ( Fig. 1a) (49,50) and likely represents an important region in the control of BK channel function. We demonstrate that a functionally diverse array of splice variants with a range of calcium and voltage sensitivities, distinct tissue distribution, differential sensitivity to protein phosphorylation, and subcellular localization are generated from a single site of splicing. The control and coordination of these splicing decisions at a single site of splicing provide a powerful mechanism to generate a physiologically diverse complement of BK channel ␣-subunits.

HEK293 Cell Culture and Transfection
HEK293 cells were subcultured essentially as described (44,51). Briefly, cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum in a humidified atmosphere of 95% air, 5% CO 2 at 37°C. Cells were passaged every 3-7 days using 0.25% trypsin in Hanks' buffered salt solution containing 0.1% EDTA. For biochemical studies, cells were grown to 70 -80% confluence in 25-or 75-cm 2 flasks. For electrophysiological or imaging assays, cells were plated on glass coverslips. Twenty four hours prior to the experiment, cells were washed, and medium was replaced with Dulbecco's modified Eagle's medium containing insulin-transferrin-sodium selenite (ITS) serum replacement (Invitrogen). Cells were transiently transfected at 40 -60% confluence with the respective cDNA using Lipofectamine 2000 (Invitrogen), essentially as described by the manufacturer. Medium was replaced after 24 h, and assays were performed 48 -72 h post-transfection. Stable cell lines were also created by selection and maintenance by using 0.8 mg/ml geneticin (Invitrogen).

Transcript Scanning for Site C2 Splice Variants
Transcript scanning was initially performed on total cDNA generated from 19-day-old whole mouse embryo RNA (embryo 19-day cDNA, mouse tissue rapid-scan cDNA pool; Origene) by PCR amplification between exons 15 and 25 (see Fig. 1b) using forward (fwd), 5Ј-GTC CTT CCC TAC TGT TTG-3Ј, and reverse (rev), 5Ј-GTG TTT GAG CTC ATG ATA GT-3Ј primers. PCR amplicons were cloned directly into the pCR2.1 TOPO vector (Invitrogen) according to the manufacturer's instructions. Clones were characterized by agarose gel electrophoresis and sequenced on both strands by automated sequencing (MWG Biotec, Germany).

BK Channel Splice Variant Expression Constructs
The cloning and subcloning of C-terminal hemagglutinin (-HA) tag constructs of the mouse BK channel splice variants ZERO and e21(STREX) in the mammalian expression vector pcDNA3 (Invitrogen) have been described previously (44,51). Full-length, -HA epitopetagged, constructs (in pcDNA3) of the e20, e22, and ⌬e23 variants and the C-terminal -GFP and N-terminal FLAG-tagged constructs, used in this study, were generated as described in the supplemental SD1.

qRT-PCR TaqMan TM Analysis
Primers and probes for TaqMan TM quantitative real time PCR (qRT-PCR) assays, specific for each murine site C2 splice variant, were designed with Primer Express version 1.2 (Applied Biosystems). Taq-Man TM probes, labeled at the 5Ј end with 6-carboxyfluorescein and at the 3Ј end with 6-carboxytetramethylrhodamine, were synthesized by Applied Biosystems.
All assays were performed using Applied Biosystems universal cycling parameters (2 min hold at 50°C, 10 min hold at 95°C, then 40ϫ (15 s at 95°C and 1 min at 60°C) cycles) on an Applied Biosystems ABI Prism 7000 sequence detection system. Reactions (25 l) were performed in ABI Prism 96-well optical reaction plates. Each reaction contained 1ϫ Applied Biosystems real time PCR master mix (including ROX passive reference dye, 5 mM MgCl 2 , and nucleotides), 50 nM each of the respective forward and reverse primers, and 5 nM of labeled Taq-Man TM probe. All data were analyzed using ABI Prism 7000 SDS software version 1.0 (Applied Biosystems). Transcript expression was determined from standard curves generated using dilutions of the respective splice variant plasmid DNA.

Total RNA and cDNA Preparation for qRT-PCR TaqMan TM Analysis
Total RNA from selected tissues was prepared using the QIAgen RNeasy mini kit according to the manufacturer's instructions. RNA was treated with RNase-free DNase, and reverse transcription was performed in 20-l reactions containing 1ϫ reverse transcriptase buffer (Qiagen), 0.5 mM of each dNTP, 1 M oligo(dT) primer or random hexamers (Amersham Biosciences), 10 units of RNasin (Promega), 4 units of Omniscript reverse transcriptase (Qiagen), and 2 g of total RNA. Reactions were incubated for 60 min at 37°C, and then cDNA products were stored at Ϫ20°C before TaqMan TM analysis. Control reactions were performed in parallel to exclude contamination from genomic DNA by exclusion of reverse transcriptase, or primers, from the reverse transcriptase reaction.

Patch Clamp Electrophysiology
Macropatch current recordings were performed in the inside-out configuration of the patch clamp technique, at room temperature (20 -24°C), using equimolar potassium gradients essentially as described (44,51). The pipette solution (extracellular) contained (in mM) the following: 140 KCl, 5 NaCl, 0.1 CaCl 2 , 2 MgCl 2 , 20 glucose, 10 HEPES, pH 7.4. The bath solution (intracellular) contained (in mM) the following: 140 KCl, 5 NaCl, 2 MgCl 2 , 30 glucose, 10 HEPES, pH 7.3. Intracellular free calcium ([Ca 2ϩ ] i ) was buffered using 1 mM BAPTA or dibromo-BAPTA as appropriate. Patches were maintained at a holding potential of Ϫ50 mV and stepped to the voltages indicated in the figure legends for 100 ms. Peak tail current amplitudes were analyzed to determine conductance-voltage curves.
Solution exchange, or application of cAMP (0.1 mM) to the intracellular face of patches, was by gravity-driven perfusion (10 volumes of the bath solution (bath volume, 0.5 ml) by gravity driven perfusion at a flow rate of 1-2 ml/min). For some experiments cAMP was added directly to the bath.
Data acquisition and voltage protocols were controlled by an Axopatch 200 A or B amplifier and pCLAMP6 or pCLAMP9 software (Axon Instruments Inc., Foster City, CA). All recordings were sampled at 10 kHz and filtered at 2 kHz. Channel activity was allowed to stabilize for at least 10 min after patch excision. BK channel activity was stable for Ͼ1 h under the recording conditions used (data not shown), in the absence of channel modulators.
Single-channel open probability (P o ) was derived from single-channel analysis using pSTAT (Axon Instruments) for patches with Ͻ4 channels. In the case of patches with Ͼ4 channels, an integration-over-baseline algorithm was used, using Igor Pro 4.1 (WaveMetrics, Lake Oswego, OR). In the latter case, N ϫ P o (number of functional channels ϫ open probability of channel) values were determined as follows. All-point histograms were plotted to obtain the "offset," i.e. leak current, as well as the single-channel current amplitude from the peak intervals. After subtraction of the offset from the traces, these were integrated over 0.5-60-s segments. The integral divided by integration time and singlechannel current amplitude gives N ϫ P o .
To determine the mean percent (%) change in channel activity after a treatment, in patches with low to moderate levels of channel expression, mean P o or N ϫ P o was averaged from 3-5 min of recording at ϩ 40 mV immediately before and 10 min after the respective drug treatment. Mean change in activity was expressed as a percentage (%) of the pretreatment control Ϯ S.E. In the respective figure legends and text a positive percentage (%) change in activity reflects activation, whereas a negative percentage (%) change reflects channel inhibition.

Immunohistochemistry and Imaging
Cell Permeabilization and Fixation-Transfected HEK293 cells were washed briefly in phosphate-buffered saline (PBS). Cells were fixed and permeabilized with 2% paraformaldehyde plus 0.3% Triton X-100 overnight at 4°C. Cells were fixed for a further 10 min at room temperature in 4% paraformaldehyde. Cells were then blocked (3% bovine serum albumin in PBS plus 0.05% Tween 20) for 1 h at room temperature. The -HA epitope tag was probed using a rabbit polyclonal anti-HA antibody (0.5 g/ml, Zymed Laboratories Inc.). After washing three times with PBS, at room temperature, cells were incubated with Alexa-488-conjugated secondary anti-rabbit antibody (1:1000, Molecular Probes). Following three washes in PBS, coverslips were mounted on microscope slides using Mowiol.
Nonpermeabilized Fluorescent Labeling-For cell surface (N-terminal FLAG epitope) labeling prior to fixation or permeabilization, a 1:100 (20 g/ml) dilution of a mouse monoclonal anti-FLAG antibody (M2, Sigma) was applied to cells for 1-2 h on ice. Alexa-594-conjugated anti-mouse IgG (1:1000, Molecular Probes) was used as the secondary antibody.
Confocal images were acquired on a Zeiss LSM510 laser scanning microscope, using a 63x oil Plan Apochromat (NA ϭ 1.4) objective lens, in multi-tracking mode to minimize channel cross-talk. All figure images are unprocessed single sections taken through the middle of the cell.

Immunoprecipitation and Western Blotting
Cells were lysed at 4°C in buffer containing 150 mM NaCl; 50 mM HEPES, pH 7.5; 1.5 mM MgCl 2 ; 10 mM sodium pyrophosphate; 20 mM NaF; 1 mM EDTA; 5 mM EGTA; 10% v/v Glycerol; 1% Triton-X-100, and proteinase inhibitor mixture (Roche Applied Science). Cell lysates were spun and supernatants pre-cleared for 30 min with protein-G beads (Sigma). Lysates were then incubated overnight at 4°C with 10 l (4.6 g/l) of anti-FLAG M2 antibody (Sigma) and 40 l of protein-G beads (1:1 in lysis buffer). Beads were washed a minimum of five times with lysis buffer. Bound complexes were eluted with SDS sample buffer, separated through a 10% SDS-polyacrylamide gel, and electroblotted to polyvinylidene difluoride membrane. Blots were probed overnight with 0.25 g/ml rabbit anti-HA polyclonal antibody (Zymed Laboratories Inc.) at 4°C before incubation with horseradish peroxidase-conjugated anti-rabbit IgG secondary antibody (1:5000, Diagnostics Scotland, UK) for 1 h at room temperature. Signals were detected using enhanced chemiluminescence, and images were captured on a Syngene Genegnome camera system using Genesnap software (Syngene).

Multiple Splice Variants Expressed at Site of Splicing C2-Transcript
scanning of cDNA generated from 19-day-old whole mouse embryo RNA, using primers designed to amplify across alternative site of splicing C1 and C2 (Fig. 1, a and b), resulted in a family of alternatively spliced transcripts with distinct inserts at site of alternative splicing C1 and C2 (Fig. 1c). Size and sequence analysis of Ͼ400 transcripts revealed expression of two distinct variants at site of splicing C1 ( Fig. 1, b and c). The insertless "null" variant and the four amino acid insert resulting from inclusion of exon 17 has been reported previously. At site C2, five distinct variants were isolated ( Fig. 1, b and c), of which four have been reported previously in the mouse as follows: the "ZERO" variant resulting from splicing of exon 19 to exon 23; the e20 variant resulting in a three-amino acid (IYF) insert from inclusion of exon 20 between exon 19 and exon 23; the e21 variant resulting in the 59-amino acid, cysteinerich, stress-regulated exon (STREX) insert resulting from inclusion of exon 21 between exon 19 and exon 23; and the e22 variant resulting in a 29-amino acid insert resulting from inclusion of exon 22 between exon 19 and exon 23. We also isolated a novel variant (⌬e23) resulting from skipping of exon 23. This results in splicing of exon 19 to exon 24 leading to a frameshift that introduces a premature stop codon within exon 24. This generates a truncated channel terminating in the amino acids . . . . RRRMTPC*. Analysis of Ͼ400 transcripts revealed that either of the null or e17 site C1 splice variants may be expressed in the same transcript with any of the site C2 splice variants (Fig. 1d).
The e21(STREX) splice variant is highly conserved, at the amino acid level, from fish to man. Nine of 59 amino acid substitutions exist between the murine and Zebrafish sequences suggestive of a common function across vertebrate evolution (Fig. 2). In contrast, the e22 splice variant is poorly conserved, with 11 of 29 amino acid substitutions between man and mouse. Thus, it is likely that distinct selection pressure and function of this variant is manifest across species.
Tissue Distribution of Splice Variant mRNA-To address the tissue distribution of each site C2 splice variant mRNA, across different mouse tissues, we developed splice variant-specific quantitative real time PCR TaqMan TM assays (see supplemental SD2). Total BK channel mRNA expression between different tissues (Fig. 3) varied over 3 orders of magnitude. Total BK channel mRNA levels were highest in adult brain, prostate, and 19-day-old whole mouse embryo. The lowest levels were observed in heart and liver. Expression of total BK channel mRNA in these tissues was less than 3% of that observed in 19-day-old whole mouse embryo. Comparison of splice variant mRNA expression across tissues, normalized to expression in 19-day-old whole mouse embryo, demonstrated tissue heterogeneity in the relative levels of each variant. Adult brain, adrenal gland, and prostate had highest levels of the ZERO variant mRNA. The e21(STREX) variant mRNA was predominantly expressed in endocrine tissues such as the adrenal gland, pituitary, and prostate. Splice variant e22 mRNA was predominantly expressed in embryonic tissue, whereas ⌬e23 mRNA levels varied by 2-fold across different tissues. In our assays, splice variant e20 was consistently below the level of detection in all tissues suggesting that this variant is expressed at less than 1% of total BK mRNA levels.
Comparison of the mRNA expression of each variant, as a proportion of total BK channel mRNA levels in individual tissues, further demonstrated that different tissues express a distinct complement of site C2 splice variants (Fig. 4). For example, in adult brain, which has one of the highest levels of total BK channel mRNA, the ZERO variant constitutes Ͼ90% of all variant mRNAs expressed. In contrast, in tissues with very low total BK mRNA levels, such as liver, the predominant splice variant mRNA is ⌬e23.
Distinct Biophysical Properties of Site C2 Splice Variants-Previous functional studies of the e21(STREX) splice variant has demonstrated a dramatic shift in the apparent calcium sensitivity of the channel compared with the ZERO variant (40,41,45). To address whether phenotypic variation in intrinsic biophysical properties is also introduced by other site C2 splice variants, we analyzed the calcium and voltage sensitivity of BK channels in macropatches from HEK293 cells expressing ). The forward (fwd) and reverse (rev) PCR primers used for transcript profiling across sites of splicing C1 and C2 hybridize to exons 15 and 25, respectively. Expanded views across site C1 and C2 illustrate constitutive and alternatively spliced exon arrangements for each splice variant described in this study. For site C1, null indicates no insert at site C1, thus junction results from splicing of exon 16 to exon 18. For site C2, ZERO indicates no insert at site C2, thus junction results from splicing of exon 19 to exon 23. These exons flank the respective exons generating the e20, e21(STREX), and e22 splice variants. For the ⌬e23 splice variant, skipping of exon 23 results in splicing of exon 19 to exon 24. c, amino acid sequence of inserts generated at sites C1 and C2 for each alternatively spliced exon. Boldface labeling indicates amino acid residues resulting from insertion of the respective exon at each site, with the three amino acids from the respective flanking exons shown. Exon 23 skipping (⌬e23 variant) results in a frameshift and generation of an in-frame stop codon, within exon 24, resulting in premature truncation of the protein (*). d, representative ethidium bromide-stained agarose gel of amplicons generated from a PCR screen of individual clones, containing distinct splice variant inserts, isolated from 19-day-old whole mouse embryo cDNA. The fragment size indicates distinct exon combinations, resulting from alternative exon usage at sites of splicing C1 and C2, as indicated.
the respective site C2 splice variant (Fig. 5). Each site C2 splice variant was expressed as a C-terminal -HA-tagged channel (Fig. 5). As reported previously, the -HA tag does not affect properties of ZERO or e21(STREX) channels, and no significant differences in channel properties compared with the un-tagged variants were observed in these studies (not shown).
Expression of ZERO, e20, e21(STREX), and e22 variants in HEK293 cells resulted in robust voltage-and calcium-sensitive out-ward macropatch currents in HEK293 cells (Fig. 5a). In contrast, no detectable macropatch currents were observed in cells transfected with the ⌬e23 splice variant, even at voltages greater than ϩ100 mV in the presence of 100 M [Ca 2ϩ ] i. This suggests that channels formed by ⌬e23 are either nonfunctional or not expressed at the plasma membrane.
The e20 splice variant displayed a calcium and voltage sensitivity similar to that previously reported for the murine ZERO variant. The  Expression of each variant, or total BK channel, mRNA is normalized to the expression in 19-day-old whole mouse embryo. Note that the e20 splice variant is expressed at Ͻ1% of transcript levels compared with total BK channel mRNA levels in 19-day-old whole mouse embryo and was below the limit of detection in these assays. Data with no bar indicate expression is Ͻ1% of embryo day 19 cDNA. Data are expressed on a logarithmic scale, means Ϯ S.E., n ϭ 3-9/group. V 0.5(max) at 1 M [Ca 2ϩ ] i was 71.2 Ϯ 6.8 and 68.3 Ϯ 5.6 mV for the e20 and ZERO variants, respectively (Fig. 5, b and c). As shown previously, inclusion of e21(STREX) results in a large left shift in the conductancevoltage curve with a V 0.5(max) at 1 M [Ca 2ϩ ] i of 11.2 Ϯ 5.4 mV (Fig. 5, b  and c). The e22 splice variant also displayed a significant left shift in the conductance-voltage relationship, compared with ZERO, with a V 0.5(max) at 1 M [Ca 2ϩ ] i of 34.3 Ϯ 8.4 mV (Fig. 5, b and c). This value is intermediate to the V 0.5(max) for ZERO and e21(STREX) under these recording conditions. Furthermore, both e21(STREX) and e22 variants had considerably slower deactivation kinetics compared with ZERO or e20 (Fig. 5a).
⌬e23 Splice Variant Is a Dominant Negative of Plasma Membrane Expression-The lack of functional expression of the novel ⌬e23 splice variant suggested that the channel is nonfunctional or is not expressed  at the plasma membrane. A conserved motif within the intracellular C terminus of mammalian BK channels is required for efficient export of channels from the endoplasmic reticulum (47), although the precise role of the intracellular C terminus in plasma membrane targeting is controversial (52). The C-terminal truncation of the ⌬e23 splice variant would result in loss of this motif. Thus, ⌬e23 channels would be predicted to reside largely in the endoplasmic reticulum. In support of this, confocal imaging of HEK293 cells, transiently or stably expressing the FLAG-and/or -HA-tagged ⌬e23 variant, revealed a predominant intracellular localization of the channel (Fig. 6a). This is in contrast to the robust plasma membrane targeting of either the FLAG-or -HA-tagged ZERO, e21(STREX), or e22 variants (Figs. 6a and 7a). The ⌬e23 splice variant was largely expressed in a diffuse perinuclear distribution and co-localized with an endoplasmic reticulum marker (not shown). The lack of membrane expression of the ⌬e23 splice variant was not a result of reduced protein expression. Indeed, both single cell image analysis (Fig. 6a) and Western blotting of cell lysates (Fig. 6b) revealed robust expression of this C-terminally truncated protein. Furthermore, inefficient membrane targeting was unlikely to be due to incorrect assembly of homomeric ⌬e23 channels as ⌬e23 splice variant ␣-subunits were able to homotetramerize (Fig. 6c). This would be in accordance with the truncation in this variant occurring after the first regulator of potassium conductance domain, which is required for ␣-subunit tetramerization (49,50).
The ⌬e23 splice variant may act as a dominant negative of cell surface expression. Conversely, ⌬e23 trafficking to the plasma membrane may be rescued upon co-assembly with a trafficking "normal" variant. To address these issues, we co-expressed the C-terminal -HA epitopetagged ⌬e23 variant (⌬e23-HA) with a C-terminal -GFP-tagged e21(STREX) variant, e21(STREX)-GFP, and we analyzed plasma membrane and perinuclear localization in HEK293 cells. Expression of ⌬e23-HA alone resulted in less than 4% of transfected cells with any visible staining at the cell periphery (Fig. 7, a and d). This is in agreement with the lack of cell surface staining observed for N-terminal FLAGtagged constructs (FLAG-⌬e23 (not shown) or FLAG-⌬e23-HA (Fig.  6a)) and the lack of functional expression of outward potassium currents in the electrophysiological assays (Fig. 5). The very low level of plasma membrane localization of the ⌬e23-HA splice variant may reflect saturation of the endoplasmic reticulum export control mechanisms. Alternatively, it may reflect an intrinsically low probability of membrane trafficking for this C-terminally truncated variant. In contrast, greater than 70% of transfected cells showed robust e21(STREX)-GFP expression at the cell periphery (Fig. 7, a and d) as for FLAG-or -HA epitope-tagged ZERO (Figs. 6a and 7d), e21 (not shown), or e22 splice variants. More importantly, co-expression of the ⌬e23-HA variant with the e21(STREX)-GFP variant completely blocked plasma membrane expression of e21(STREX)-GFP in HEK293 cells. Co-expression resulted in co-localization of e21(STREX)-GFP with ⌬e23-HA in a perinuclear-endoplasmic reticular localization identical to that of ⌬e23-HA alone (Fig. 7, a and d). Identical intracellular trapping was also observed on co-expression of the ⌬e23 variant with ZERO or e22 splice variants (not shown). This strongly suggests that the ⌬e23 splice variant is a dominant negative of cell surface expression.
The enhanced intracellular trapping of e21(STREX), in the presence of ⌬e23, is most likely because of co-assembly of the ⌬e23 and e21(STREX) splice variant ␣-subunits as heteromultimers in the endoplasmic reticulum. In support of this conclusion, epitope-tagged ⌬e23 and e21(STREX) variants could be reciprocally co-immunoprecipitated from HEK293 cells (Fig. 7, b and c). Trafficking of the e21(STREX)-GFP splice variant to the plasma membrane was not affected by co-expression with the ZERO-HA variant (Fig. 7, a and d). Thus, the lack of plasma membrane targeting, and enhanced perinuclear localization, of the e21(STREX)-GFP variant in the presence of ⌬e23-HA was not simply an artifact of overexpression of the two variants.
As a further functional test of whether heteromultimerization of ⌬e23 with other splice variants may generate functional channels at the plasma membrane, we analyzed single channel events in isolated insideout patches of HEK293 cells co-expressing the ⌬e23 variant with either ZERO or e21(STREX) variants. For these assays, we exploited ZERO and e21(STREX) splice variant mutants (Y294V-HA and STREX-Y294V-HA, respectively) that contain a single point mutation (Y294V) in the channel pore to reduce sensitivity to extracellular tetraethylammonium (TEA) (51). If the TEA-sensitive ⌬e23 ␣-subunits assemble with ␣-subunits of another (TEA-insensitive) splice variant, and form functional channels at the plasma membrane, we would predict that FIGURE 6. Deficiency in plasma membrane trafficking of the novel ⌬e23 splice variant. a, representative single confocal sections through the center of HEK293 cells transiently expressing epitope-tagged ZERO, e22, or ⌬e23 splice variants. Variants expressed an extracellular N-terminal -FLAG tag and an intracellular C-terminal -HA tag. Surface BK channel expression was determined by exposing cells to ␣-FLAG antibodies (surface) prior to fixation and permeabilization. Total channel expression was determined following cell permeabilization with Triton X-100 and probing with ␣-HA antibodies. Images are shown as separate channels for surface (FLAG, green) and permeabilized (-HA, red) with images merged in the right-hand side panels. Note robust cell surface staining for ZERO and e22 variants but not the ⌬e23 variant. Scale bars are 5 m. Lack of cell surface expression of ⌬e23 is not because of reduced channel expression (b) or the inability to tetramerize (c). Representative Western blots of HEK293 cell lysates expressing e22 and ⌬e23 variants separated by SDS-PAGE under reducing (10% polyacrylamide gel) (b) or nonreducing (5% polyacrylamide gel) (c) conditions. c, the number of * indicates predicted size of monomer, dimer, trimer, and tetramer, respectively. Channel expression in immunoblots was probed using ␣-HA antibodies and enhanced chemiluminescence. single channel events, with a reduced single channel amplitude in the presence of 2 mM extracellular TEA, would be observed (51). The single channel amplitude under these conditions is an indication of the stoichiometry of the TEA-sensitive (⌬e23 variant) to the TEA-insensitive (Y294V-HA or STREX-Y294V-HA variant) ␣-subunits in the tetramer. We only observed one patch (out of more than 120 patches exposed to greater than ϩ80 mV and 100 M [Ca 2ϩ ] i ) in which a large conductance channel was observed, when the ⌬e23 variant was expressed alone in HEK293 cells. This is in agreement with the lack of ⌬e23 splice variant protein expression at the plasma membrane (Figs. 6 and 7). In no patch (out of more than 50 tested) were any BK channels observed that were sensitive to 2 mM external TEA when ⌬e23 was co-expressed with another, TEAinsensitive, splice variant (not shown). Thus, all single channel events recorded in these co-expression assays were the result of homotetramers of the respective TEA-insensitive ZERO, or e21(STREX), variants.
Taken together, these data support the hypothesis that the novel ⌬e23 splice variant acts as a dominant negative of cell surface expression.
Regulation of Splice Variants by PKA-dependent Phosphorylation-The single channel conductance of each variant was determined in excised inside-out patches exposed to physiological potassium gradients in which less than four channels were present. There was no significant difference in single channel conductance between the ZERO, e21(STREX), and e22 splice variants (121.7 Ϯ 8.9, 123.8 Ϯ 5.4, and 125.2 Ϯ 7.4 pS, respectively). In contrast, the e20 splice variant had a significantly increased single channel conductance, under identical recording conditions, of 143.3 Ϯ 7.2 pS. In only one patch (out of more than 120 patches exposed to greater than 100 M [Ca 2ϩ ] i and depolarized to greater than ϩ80 mV) did we observe a single channel event in HEK293 cells transfected with the ⌬e23 splice variant. Although full characterization was not possible, the single channel conductance was similar to that of the other variants (117 pS). We have never recorded such a channel in our native or mock-transfected HEK293 cells under these recording conditions. However, this channel was only active at voltages greater than ϩ60 mV in the presence 100 M [Ca 2ϩ ] i .
Alternative pre-mRNA splicing is an important determinant of BK channel regulation by PKA phosphorylation (22,43,44). As we have reported previously, ZERO splice variants are activated (mean activation was 110.1 Ϯ 9.5%, n ϭ 8 of control; Fig. 8, a and b) by closely associated PKA, in excised inside-out patches, upon application of 0.1 mM cAMP to the intracellular face of the patch. In contrast, under identical recording conditions, the e21(STREX) variant is inhibited by endogenous PKA (mean inhibition was Ϫ60.6 Ϯ 10.3%, n ϭ 5 of control; Fig. 8, a and b). The effect of cAMP on both ZERO and e21(STREX) channel activity is completely abolished in the presence of the specific PKA inhibitor peptide, PKI-(5-24), or in the absence of ATP. The mean cAMP-induced change in activity in the presence of PKI-(5-24) was Ϫ2.1 Ϯ 4.5%, n ϭ 4, and 1.3 Ϯ 2.4%, n ϭ 4, for ZERO and e21(STREX), respectively. In the absence of ATP, the mean change in activity was 1.1 Ϯ 6.2%, n ϭ 4, and 2.1 Ϯ 4.0%, n ϭ 4, respectively.
To examine whether introduction of other alternatively spliced inserts, at site of splicing C2, also modifies BK channel regulation by PKA-dependent protein phosphorylation, we assayed the e20 and e22 splice variants under identical conditions. Both the e20 and e22 splice variants displayed activation by endogenous PKA similar to that observed with the ZERO variant. Mean activation in the presence of cAMP was 128.6 Ϯ 25.4%, n ϭ 10, and 76.8 Ϯ 12.7%, n ϭ 6, for the e20 and e22 splice variants, respectively. More importantly, the effect of cAMP on both variants was completely abolished in the presence of PKI-(5-24) (mean change in cAMP-induced activity was Ϫ7.3 Ϯ 8.2%, n ϭ 6, and Ϫ5.2 Ϯ 5.6%, n ϭ 6, for e20 and e22, respectively). Similarly, the effect of cAMP was completely abolished in the absence of ATP (mean change in activity was 2.9 Ϯ 5.6%, n ϭ 4, and Ϫ0.8 Ϯ 11.8%, n ϭ 4, for the e20 and e22 splice variants, respectively). Thus, the e20 and e22 splice variants are activated in a similar manner to the ZERO variant by PKA phosphorylation. The e21(STREX) splice variant, which introduces an additional PKA consensus motif (44) within the STREX insert, is the only site C2 splice variant to be inhibited by PKA.

DISCUSSION
The pore-forming ␣-subunits of BK channels are encoded by a single gene (KCNMA1) (28 -30). Considerable functional variation may be generated by alternative pre-mRNA splicing in species as diverse as worms and humans. Inclusion of specific exons is known to modify physiologically relevant properties of the expressed channel protein. However, the extent to which exon inclusion, or exclusion, at a single site of splicing is important in generated functional diversity of BK channels is poorly understood.
In this report, we have isolated and systematically characterized the functional properties of five, alternatively spliced, murine BK channel variants that are generated by splicing at site C2 (see Fig. 1a). Four of the splice variants have been isolated previously (ZERO, e20, e21(STREX), and e22), although only the ZERO and e21(STREX) splice variants have been characterized functionally. We have also identified a novel, C-terminally truncated, splice variant (⌬e23) that results from skipping of exon 23. This exon was previously thought to be a constitutively expressed exon.
Individual splice variants show distinct patterns of tissue distribution, cell surface expression, biophysical properties, and regulation by cAMPdependent protein phosphorylation. Taken together, alternative splicing at site of splicing C2 would provide a powerful mechanism to specify BK channels with a spectrum of properties. These channels may be individually tailored to the properties of individual cell types, perhaps at different developmental stages or in response to a physiological challenge.
Differential mRNA Expression of Functionally Diverse Splice Variants in Murine Tissues-We have developed high sensitivity and specificity TaqMan TM real time PCR assays to quantitate, for the first time, the mRNA expression of each site C2 splice variant in different murine tissues. By using these assays, total BK channel mRNA levels were found to be highest in adult brain and a number of endocrine tissues but with extremely low mRNA expression in heart and liver. These total BK channel mRNA expression patterns largely agree with previous Northern blots, as well as protein and functional expression studies in other mammalian species.
Considerable diversity in splice variant mRNA expression profile, across different tissues, was observed. This may provide fine tuning of BK channel properties in different tissues according to physiological need. We cannot address directly whether the mRNA expression profiles determined for each splice variant directly translates into levels of protein (or functional) channel in each tissue. However, the mRNA expression profile likely provides an important correlate of splice variant protein expression. For example, e21(STREX) variant mRNA is highly expressed in endocrine tissues such as the adrenal gland and pituitary, as described previously (40,41), whereas ZERO is the predominant brain variant. Heterologously expressed e21(STREX) variant channels display a significant left shift in half-maximal voltage of activation (V 0.5(max) ) compared with ZERO (40,41,45,53). Furthermore, e21(STREX) variants are also potently inhibited by cAMP-dependent protein kinase phosphorylation (44,45), whereas ZERO is activated. In many anterior pituitary cell types, native BK channels display a leftshifted V 0.5(max) and are potently inhibited by PKA phosphorylation (10,45,54), properties that correlate with the STREX phenotype. In contrast, many neuronal BK channels reveal properties similar to the ZERO variant: low calcium sensitivity and activation by protein kinase A. The native BK channels in brain largely display a lower calcium sensitivity and are activated by PKA, properties consistent with a ZERO phenotype. Further detailed expression and functional analysis is required to address the spatial and temporal expression of BK channel splice variants at the mRNA, protein, and functional channel level in different tissues.
The e21(STREX) insert is highly conserved across vertebrates, at the amino acid level, suggesting it plays a common functional role, perhaps in the maintenance of high frequency or duration of action potentials (11,20). To date, a homologous e21(STREX) variant in invertebrate species (such as Drosophila or Caenorhabditis elegans) has not been identified; thus e21(STREX) likely represents a vertebrate specialization.
In contrast to the e21(STREX) splice variant, the e22 variant is poorly conserved between mouse and human (11 of 29 amino acid substitutions), indicating that the properties, and function, of this variant may be distinct between species. Indeed, inclusion of the human e22 variant results in no significant shift in the conductance-voltage relationship (55). This is in contrast to the significant (greater than 20 mV) left shift of the V 0.5(max) conferred by the murine e22 variant in this study. The V 0.5(max) and rate of channel deactivation was intermediate to that of the ZERO and e21(STREX) splice variants. Furthermore, although the e22 splice variant displays biophysical characteristics intermediate to that of e21(STREX) and ZERO, the response to protein kinase A phosphorylation is identical to that of ZERO channels. As the ZERO, e20, and e22 variants retain the conserved C-terminal PKA consensus phosphorylation motif, it is likely that this site is important for mediating PKA-dependent activation in all three variants as is the case for ZERO (44,51). Transcripts containing exon 22 have been described previously in cDNA libraries from murine brain and skeletal muscle (29). The homologous exon is expressed in human (55) and rat (56) brain, and a long variant is up-regulated in human glioma tissue (39). However, in the murine tissues examined, expression of the e22 splice variant mRNA is relatively low in adult tissues apart from breast. In contrast, the e22 splice variant mRNA is expressed at high levels at embryonic stages in the mouse. Whether this represents a developmentally restricted pattern of expression or, alternatively, expression in tissues not sampled in the adult in our assays remains to be determined. Taken together, expression of these distinct variants would provide an array of BK channels whose functional impact would be optimal across a wider dynamic range of cellular voltages, intracellular free calcium levels, and in response to modulation of intracellular signaling cascades.
A Novel C-terminal Splice Variant Dominant Negative ␣-Subunit of Cell Surface Expression-We have isolated and characterized a novel splice variant (⌬e23) that results from skipping of exon 23. The ⌬e23 splice variant is deficient in trafficking to the cell surface and acts as a dominant negative inhibitor of cell surface expression of other splice variants. This phenotype is similar to a C-terminally truncated variant in rabbit, rbSlo2 (47), generated as result of splicing at a site 3Ј to site of splicing C2. The intracellular trapping of the ⌬e23 splice variant may thus be because of the lack of the recently identified endoplasmic reticulum exit signal, located between residues 1105 and 1110, in full-length channels (47). As for rbSlo2, the murine ⌬e23 splice variant can assemble as tetramers. However, we observed no significant decrease in total protein expression of the ⌬e23 variant, compared with full-length channels, as reported for rbSlo channels that are truncated N-terminal to amino acid Asn-1061 (47). In humans, insertion of an alternatively spliced exon at the N terminus also generates BK channels with defective trafficking to the cell surface. These ␣-subunits also act as a dominant negative of cell surface expression (48). Thus, exon inclusion (48) as well as exon exclusion (this study and Ref. 47) can generate BK channel ␣-subunits with deficient cell surface trafficking.
Is there a physiological function for the ⌬e23 and other traffickingdeficient splice variants? In the mouse myometrium, total BK channel protein expression increases toward term. Paradoxically, the BK channel current decreases (25,57). In this system, cell surface clustering of BK channel ␣-subunits is reduced with redistribution to a perinuclear/ endoplasmic reticulum localization (57). Whether up-regulation of the ⌬e23 splice variant, in late pregnancy, underlies such down-regulation of cell surface expression remains to be determined. Furthermore, in tissues with low total levels of BK channel mRNA expression, the ⌬e23 variant is the predominant splice variant expressed. This suggests that transcriptional as well as post-translational mechanisms are used, in these tissues, to suppress functional BK channel expression.
How the splicing decisions, for the alternatively spliced exons at site of splicing C2, are controlled and coordinated are poorly understood (27,58). Alternative splicing of the e21(STREX) exon can be dynamically regulated, for example, in response to circulating hormones (24,27), during stress (23), pregnancy (25), and upon cellular depolarization (58). However, factors that may regulate splicing decisions for the e20, e22, or ⌬e23 splice variants are essentially unknown. An intronic calcium-calmodulin regulatory response element has been identified as an important factor in control of e21(STREX) exon splicing. However, similar elements do not control splicing of exon 22 (58).
Taken together, alternative splicing at site of splicing C2 would provide a powerful mechanism to specify BK channels with a spectrum of properties. This would include differences in calcium sensitivity, kinetics, cell surface expression, and regulation by protein phosphorylation at the plasma membrane. As the splicing decision can be dynamically regulated, this would allow channel properties to be individually tailored to the demands of individual cell types. This may be manifest at different developmental stages or in response to a physiological challenge. Such functional diversity, and plasticity, generated by splicing at a single site of pre-mRNA splicing, may be further enhanced, for example, by coordination of splicing at other sites of splicing, heteromultimerization of ␣-subunit splice variants and/or assembly with distinct regulatory ␤-subunits. Characterization of the mechanisms controlling splicing decisions at the site of alternative splicing C2 should provide important insights into the functional role of these phenotypically distinct variants. Importantly, our data reveal new insights into the diversity, and plasticity, of the functional properties of BK channels that can be generated by alternative pre-mRNA splicing from a single site of splicing.