Originally published as Biophys J. BioFAST on August 12, 2005.
doi:10.1529/biophysj.105.063610
Biophysical Journal 89:3079-3092 (2005)
© 2005 The Biophysical Society
Homology Modeling Identifies C-Terminal Residues that Contribute to the Ca2+ Sensitivity of a BKCa Channel
Jian-Zhong Sheng *,
Aalim Weljie 
,
Lusia Sy *,
Shizhang Ling *,
Hans J. Vogel and
Andrew P. Braun *
Departments of * Pharmacology and Therapeutics and Biological Sciences, University of Calgary, Calgary, Alberta, Canada
Correspondence: Address reprint requests to Andrew P. Braun, Dept. of Pharmacology and Therapeutics, Faculty of Medicine, University of Calgary, 3330 Hospital Dr., NW, Calgary, Alberta T2N 4N1, Canada. Tel.: 403-220-8861; Fax: 403-270-2211; E-mail: abraun{at}ucalgary.ca.
 |
ABSTRACT
|
|---|
Activation of BKCa channels by direct Ca2+ binding and membrane depolarization occur via independent and additive molecular processes. The "calcium bowl" domain is critically involved in Ca2+-dependent gating, and we have hypothesized that a sequence within this domain may resemble an EF hand motif. Using a homology modeling strategy, it was observed that a single Ca2+ ion may be coordinated by the oxygen-containing side chains of residues within the calcium bowl (i.e., 912ELVNDTNVQFLD923). To examine these predictions directly, alanine-substituted BKCa channel mutants were expressed in HEK 293 cells and the voltage and Ca2+ dependence of macroscopic currents were examined in inside-out membrane patches. Over the range of 1–10 µM free Ca2+, single point mutations (i.e., E912A and D923A) produced rightward shifts in the steady-state conductance-voltage relations, whereas the mutants N918A or Q920A had no effect on Ca2+-dependent gating. The double mutant E912A/D923A displayed a synergistic shift in Ca2+-sensitive gating, as well as altered kinetics of current activation/deactivation. In the presence of 1, 10, and 80 mM cytosolic Mg2+, this double mutation significantly reduced the Ca2+-induced free energy change associated with channel activation. Finally, mutations that altered sensitivity of the holo-channel to Ca2+ also reduced direct 45Ca binding to the calcium bowl domain expressed as a bacterial fusion protein. These findings, along with other recent data, are considered in the context of the calcium bowl's high affinity Ca2+ sensor and the known properties of EF hands.
|
INTRODUCTION
|
|---|
The large conductance, calcium-activated potassium channel (BKCa channel) contributes to stimulus-induced changes in membrane potential in neurons, smooth muscle, and many secretory cells through its ability to open in response to membrane depolarization and elevated cytosolic free calcium in an additive fashion. The predicted membrane topology of the pore-forming BKCa 
-subunit closely resembles those of voltage-gated K+ channel subunits (e.g., Kv1-4 family members), with the added features of an extracellular N-terminus and long intracellular C-terminal tail. Initial efforts by Salkoff and co-workers to unravel the molecular properties underlying the BKCa channel's calcium sensitivity have shown that a series of C-terminal Asp acid residues, dubbed the "calcium bowl", appears to be critically involved (31
,32). Subsequent studies taking advantage of mutational analyses have identified two additional regions within the large C-terminus of the BKCa -subunit that selectively influence the calcium sensitivity of channel activation (3,38). Thus, neutralization/elimination of acidic residues in the calcium bowl, together with substitutions at positions D362 and D367 (e.g., D362A/D367A) (38) or M513 (e.g., M513I) (3) in the C-terminus have been shown to abolish the calcium-dependent activation of BKCa channels at free Ca2+ concentrations 
100 µM. An additional low-affinity divalent metal binding site, located within the channel's "RCK domain" (34,38), further contributes to activation at concentrations of divalent cation >0.1 mM. Very recently, Lingle and co-workers (39) have demonstrated that the regulation of BKCa channel gating by each of these three sites displays a distinct activation profile in response to a series of divalent metals (e.g., Ca2+, Sr2+, Cd2+, Mn2+, Co2+, and Ni2+), thereby strengthening the argument that the BKCa -subunit contains three physically separate divalent metal binding sites. Such activation profiles may further suggest differences in the structural features underlying each of these sites.
Using a 45Ca overlay assay, it has been demonstrated that a 
230 amino acid fragment of the BKCa -subunit from either mouse brain (2,5) or Drosophila (4) directly binds radioactive 45Ca and that this binding is disrupted by replacement of acidic residues within the calcium bowl segment. Such observations are consistent with, but do not prove, the hypothesis that this domain directly contributes to high-affinity Ca2+ ion binding, rather than acting as a "transduction module" that transfers the energy of Ca2+ binding at a distinct site to the gating machinery in the channel core. Although it is now evident that the calcium bowl functionally contributes to BKCa channel activation by physiologic levels of cytosolic Ca2+, the structural features of this domain that underlie Ca2+ binding remain undefined. Having noted earlier that the positions of oxygen-containing residues within this region align similarly with those in a canonical EF hand motif (5), we have utilized a homology modeling strategy to generate a predicted structure of this domain that describes the coordination of a single Ca2+ ion by acidic amino acids within a putative loop structure. After experimental examination of these predictions, our results show that alanine substitution of acidic residues, corresponding to the X and –Z binding ligands within a canonical EF hand, decreased the calcium sensitivity of BKCa channel activation in a synergistic manner by micromolar concentrations of Ca2+. These same substitutions also significantly decreased direct 45Ca binding to this domain expressed as a fusion protein in bacteria. In light of these data and other recent reports, we consider the possibility that an EF hand-like motif may contribute to the properties of the high-affinity Ca2+ sensor located within the calcium bowl region of the BKCa -subunit.
|
MATERIALS AND METHODS
|
|---|
The cDNA constructs encoding the mouse brain mSlo -subunit (26) and green fluorescent protein, along with procedures for site-directed mutagenesis and transient transfection of HEK 293 cells have been recently described (5). All mutations were directly confirmed by either single or double-stranded cDNA sequencing. Note that the amino acid numbering of our mSlo clone, which contains a C-terminal splicing insert (15), is slightly different than that used by other investigators for another mouse brain mSlo clone without the insert (6). As a consequence, identical residues in the calcium bowl that have been mutated by both us and other investigators (e.g., Cox and colleagues) are denoted by higher numbering in our study. To assist the reader, we have indicated identical amino acids using both numbering schemes in the text, where appropriate.
Electrophysiological measurements
Macroscopic currents were recorded at 35 ± 0.5°C from excised inside-out membrane patches of HEK 293 cells transiently transfected with either wild-type or mutant BKCa channels, as previously described (5). For mutant channels with reduced calcium sensitivity (e.g., E912A/D923A), a more positive range of voltage-clamp steps was often used to produce maximal open probability of the macroscopic currents. Micropipettes were filled with a solution containing (in millimolar): 5 KCl, 125 KOH, 1 MgCl2, 1 CaCl2, 10 HEPES, pH adjusted to 7.3 with methanesulfonic acid and had tip resistances of 1.5–3.5 M
. Voltage clamp error associated with series resistance was compensated to a level of 80% using the electronic circuitry of the Axopatch 200B amplifier. The bath solution contained (in millimolar): 5 KCl, 125 KOH, 1 MgCl2, 2 EGTA or HEDTA, 10 HEPES; the pH was adjusted to 7.25 with methanesulfonic acid. The level of free calcium in each bath solution was then independently confirmed using a calcium electrode (Orion model 93-20) with calibration standards (World Precision Instruments, Sarasota, FL) ranging from pCa 8 to 2.
For recordings in the presence of elevated MgCl2, the bath solution was modified to account for changes in osmolality with increasing concentrations of MgCl2. The 1-mM MgCl2 solution contained 50 mM KOH and 80 mM N-methyl D-glucamine (NMDG), the 10 mM MgCl2 solution contained 50 mM KOH and 70 mM NMDG, and the 80 mM MgCl2 solution contained only 50 mM KOH. The desired level of free calcium in each solution was obtained by addition of either 2 mM EGTA or 100 µM CaCl2, and pH was adjusted to 7.2 with methanesulfonic acid. The recording pipette was filled with the 1-mM MgCl2 bath solution (see above) that was supplemented with 100 µM CaCl2. Although intracellular NMDG has been previously reported to produce a modest block of BKCa channel current (21), the effect should be minimal at the tail current potentials used in our experiments (i.e., –80 and –120 mV). Furthermore, any amount of block would be expected to be uniform for both the wild-type and mutant channels at these tail potentials and should therefore not affect either the calculated half-maximal voltages of activation or the comparisons made between the channel types at 1 and 10 mM free Mg2+ concentrations.
Computer modeling of the BKCa channel C-terminal domain
The structural coordinates corresponding to a 36-amino-acid sequence spanning the third EF hand (EFIII) of human calmodulin were extracted from the Protein Data Bank file 1CLL.pdb and used as a template for an equivalent stretch of residues from the C-terminus of the BKCa channel -subunit. Model building was performed using the Homology Modeling module of the software package Insight II (Accelrys, San Diego, CA), and the step-by-step modeling procedure closely followed that recently utilized to generate an experimentally valid model of the calmodulin-like domain of a calcium-sensitive protein kinase from soybean (36). Provided in this study is a detailed description of the structural and energetic constraints invoked as the modeling procedure progressed toward the determination of a final predicted structure (36). The procedure consisted of iterative rounds of molecular dynamics and energy minimization differentially applied to regions of the calcium bowl considered to be either "variable" or "conserved". The parameters chosen for modeling the selected BKCa channel C-terminal sequence were sufficiently stringent to indicate if some aspect of the modeled sequence was grossly inadequate (e.g., too much steric interference by large residues, the length of the sequence was too long/short to act as an EF hand, etc.) and would prevent the generation of a plausible model. In addition, as we reasoned a priori that an EF hand-like motif may exist within the calcium bowl, the regions postulated to be part of the EF hand motif were treated as "conserved" during the modeling process and the remainder as "variable". Charge effects were taken into account by incorporating a single Ca2+ ion in the calculations used to generate the final predicted structure. The software program MolMol was used to create the final images shown in Fig. 2.
View larger version (37K):
[in this window]
[in a new window]
|
FIGURE 2 Structural models of the calcium bowl region generated by homology modeling using atomic coordinates from the Ca2+-bound form of human calmodulin (7). In panel A, the backbone carbon structure of the third EF hand of calmodulin is shown in a ribbon diagram; the side chains of the six Ca2+ coordinating amino acids are shown in ball-and-stick format, with oxygen, nitrogen, and carbon atoms represented by red, blue, and black balls, respectively. Note that the same color-coding scheme is also used for these atoms in panels B–F. Hydrogen atoms have been omitted for clarity. The position of a single bound Ca2+ ion within the EF hand structure is represented by the green sphere. Panel B shows a molecular model generated using a 36-amino-acid sequence of a mouse brain BKCa -subunit (26) that contains potential Ca2+ coordinating residues with homology to those found within an EF hand (refer to Fig. 1). The oxygen-containing side chains of these amino acids (i.e., E912, N915, D916, N918, Q920, and D923) are denoted by asterisks both in the model and the single letter, primary sequence shown below. Also displayed are the oxygen-containing side chains of the five Asp residues within the calcium bowl that have been previously reported to influence the Ca2+ sensitivity of BKCa channel activation (2,4,31,32,38). Note that all of the residues with oxygen-containing side chains shown in the model are indicated by red lettering in the primary sequence beneath. Shown in panels C–F are models of BKCa -subunit mutants generated after alanine replacement of one or more acidic residues in the channel's primary sequence. For clarity, the side chains of the five Asp residues depicted in panel B have been omitted. The green sphere within each predicted BKCa channel structure represents a bound Ca2+ ion, the position of which has been assigned by the modeling software in accordance with energetic constraints.
|
|
Direct 45Ca binding to bacterial fusion proteins
A region of 148 amino acids (i.e., 861NSPV.. to ..CRVA1009) that included the entire calcium bowl domain was isolated as a cDNA fragment by polymerase chain reaction from the C-terminus of the wild-type BKCa -subunit, and mutant -subunits containing engineered amino acid substitutions. Individual cDNA fragments were subcloned into the bacterial expression vector pET15b using NdeI and BamHI restriction sites, which resulted in the addition of a 21-amino-acid segment containing a 6x His affinity purification tag to the N-terminus of the BKCa channel fragment. Expression of His-tagged fusion proteins in the BL21 (DE3) strain of E. coli (200-ml cultures) was induced for 3 h at 35°C after addition of 0.1 mM IPTG. Bacteria were collected by centrifugation at 1000 x g for 10 min at 4°C and then resuspended in buffer A (i.e., 50 mM NaH2PO4, 300 mM NaCl, pH 7.0) containing 1 mg/ml lysozyme. Bacteria were lysed by freeze/thaw, sonicated for 1 min, and the BKCa -subunit fusion proteins were extracted from the insoluble material by a 1-h incubation at room temperature in buffer B (10 mM Tris HCl, pH 8, 100 mM NaH2PO4, 6 M urea, and 10 mM imidazole). After incubation of solubilized fusion proteins with 1–2 ml of Ni2+-NT agarose beads (Qiagen, Valencia, CA) (1 h, 4°C), the resin was washed three times with 20 mM Tris HCl, pH 8, 500 mM NaCl, 6 M urea, and 40 mM imidazole. Bound proteins were then eluted with buffer containing 50 mM NaH2PO4, 300 mM NaCl, and 250 mM imidazole, pH 8. After overnight dialysis to remove excess salts, equal amounts of purified fusion proteins were mixed with Laemmli sample buffer, resolved by SDS-PAGE, and then electrotransferred to 0.2 µm nitrocellulose membrane. Direct 45Ca binding to proteins transferred to nitrocellulose membrane was performed essentially as previously described (5). Briefly, nitrocellulose membranes were washed four times in buffer (60 mM KCl, 5 mM MgCl2, 10 mM imidazole HCl, pH 6.8) and then incubated for 10 min at room temperature with the same buffer containing 10 µM 45Ca. Membranes were washed for 5 min in 200 ml of 50% ethanol to remove unbound 45Ca and then dried. 45Ca binding to individual fusion proteins was quantified using a PhosphorImager (Molecular Dynamics, Piscataway, NJ) and membranes were then stained with 0.1% (w/v) amido black. The amount of fusion protein on the membrane corresponding to each 45Ca-labeled band was quantified using Quantity One image analysis software (Bio-Rad Laboratories, Hercules, CA).
Data analysis
Pairs of current-voltage relations were recorded at each concentration of free Ca2+, and the current families were then averaged for subsequent analysis. Normalized conductance-voltage (G-V) relations were calculated from tail current amplitudes measured 0.25 ms following the step to the tail potential. All G-V relations were fit with single Boltzmann functions, according to the equation:
 | (Eq. 1) |
where Vm is the experimental test potential (in volts), V1/2 is the half-maximal voltage of activation (in volts), defined as the membrane potential at which 50% of the channels are open, z is the valence of the permeable ion, and F/RT = 37.67/V at 35°C.
The time constants (
) of deactivation were derived from single exponential fits of macroscopic tail currents recorded at potentials ranging from –100 to +50 mV, in response to activating voltage-clamp steps to either 160 mV (4 µM Ca2+) or 80 mV (10–400 µM Ca2+). Exponential fits were performed over a period of 4–5 ms, beginning 0.25–0.3 ms after the step to the tail current potential. Similarly, time constants of activation were determined from single exponential fits to currents over a period of 7–10 ms, beginning 0.25 ms after the start of the voltage-clamp step to the indicated membrane potentials.
Values for wild-type and mutant BKCa channels were examined statistically using either a one-way analysis of variance and appropriate post-hoc test or an unpaired Student's t-test; differences between values were considered to be statistically significant at level of p < 0.05.
|
RESULTS
|
|---|
Sequence alignment of the calcium bowl region with an EF hand motif raises the possibility of structural similarity
We have previously reported that a series of residues contained within the calcium bowl region of the BKCa channel -subunit displays characteristics reminiscent of the Ca2+-binding properties of an EF hand structure (5). As noted in Fig. 1, the amino acid sequence adjacent to a series of aspartic acids contains a number of residues with oxygen-containing side chains in positions that correspond to the Ca2+ coordinating ligands present in the divalent ion binding loops of both classic and variant EF hand structures (19). In this study, we have utilized a homology modeling strategy to examine more objectively whether this region associated with the calcium bowl domain may be predicted to resemble a bona fide EF hand and if additional residues contributing to the calcium sensitivity of channel gating via this domain could be thus identified.
Fig. 2 A shows the solved structure of the third EF hand of human calmodulin in its Ca2+-bound form (7). The helix-loop-helix arrangement of the main-chain carbon backbone is evident, along with the single Ca2+ ion coordinated by side-chain oxygen atoms near the center of the loop. For purposes of clarity, only the residues that serve as direct Ca2+-coordinating ligands are displayed. Using standard homology modeling techniques, this EF hand structure was utilized as a template to model an analogous series of 36 amino acids within the calcium bowl region of the BKCa channel -subunit (refer to sequence alignment in Fig. 1). The -carbon backbone of the predicted structure (Fig. 2 B) displays a helix-loop-helix pattern, with a less well-ordered region beyond Asp-929 (the last side-chain group displayed in the C-terminal region). Of the six residues with oxygen-containing side chains in the sequence alignment that may act as potential Ca2+-coordinating ligands (i.e., Glu-912, Asn-915, Asp-916, Asn-918, Gln-920, and Asp-923) (refer to Fig. 1), the side chains of only Glu-912, Asp-916, and Asp-923 appear to be oriented toward the center of the predicted loop structure, such that they may physically coordinate a single Ca2+ ion (Fig. 2 B). The predicted distances between these side-chain oxygen atoms and the bound Ca2+ ion range from 2.2 to 3 Å, which are consistent with those seen in the loop structures of 4Ca2+-bound form of calmodulin (13). As the side chains of the three remaining oxygen-containing residues (i.e., Asn-915, Asn-918, and Gln-920) are positioned away from the center of the loop, it may be anticipated that they would not contribute to Ca2+ ion binding. In addition, the model shows the side chains of the five adjacent Asp residues, which appear to form an -helical turn. Previously, it has been demonstrated that replacement/deletion of some or all of this Asp-rich sequence dramatically reduces the Ca2+ sensitivity of BKCa channel gating (2,3,31,32,38). In our model, several of these side chains are oriented toward the bound Ca2+ ion, suggesting that they may influence Ca2+ binding. Bao and colleagues have recently shown that a stretch of amino acids within the calcium bowl encompassing this Asp-rich region (i.e., QDDDDDPDTELY) may be modeled as a Ca2+-binding structure when the 12-residue loop region of the first EF hand of carp parvalbumin (i.e., QDKSGFIEEDEL) is used as a structural template (2).
Given the predicted roles of Glu-912, Asp-916, and Asp-923 as Ca2+-binding ligands in the wild-type BKCa -subunit, we also generated homology models containing single or double alanine substitutions of these predicted Ca2+-coordinating residues by repeating the modeling procedure with the modified primary amino acid sequences (Fig. 2, C–F). We did not observe significant changes in the main-chain carbon backbone structure of these mutant models, indicating that these individual substitutions did not energetically destabilize the modeling routine.
Replacement of predicted Ca2+-binding residues decreases the calcium sensitivity of steady-state gating and the kinetics of current activation/deactivation
To test experimentally the predicted importance of the Ca2+-binding residues identified by our modeling efforts on the calcium sensitivity of BKCa channel gating, we carried out individual alanine substitution of the six oxygen-containing residues noted above and transiently expressed the mutant BKCa channel -subunits in HEK 293 cells. Several of the mutations examined in our study have also been recently described by Bao and colleagues using another mouse brain mSlo clone (2) that lacks a C-terminal splice insert (15). To allow the reader to follow more easily the two data sets, we have included their numbering scheme (in parentheses) along with ours when referring to the same amino acids.
Macroscopic current activities of both wild-type and mutant channels were recorded in excised inside-out membrane patches in the absence and presence of free cytosolic Ca2+. Fig. 3 shows representative tracings of current families recorded over a broad range of membrane voltage and cytosolic [Ca2+]s. Compared to the wild-type channel, currents arising from the E912(884)A and D923(895)A mutants displayed slower activation and faster deactivation kinetics at free [Ca2+]s ranging from 1 to 100 µM, and these characteristics were more pronounced in the E912A/D923A double mutant. However, at higher Ca2+ concentrations (i.e., 400 and 1000 µM), kinetic changes observed in either the single or double mutant channels were no longer evident. In contrast to these observations, alanine substitution of either D916(888) (Fig. 3), N918(890), or Q920(892) (data not shown) did not produce obvious changes in the calcium-dependent gating of these channels. Unfortunately, the N915(887)A mutant did not express well enough in our hands for reliable measurements; however, this same mutation appeared to have no effect on the calcium sensitivity of another mouse brain BKCa channel expressed in Xenopus oocytes (2).
View larger version (42K):
[in this window]
[in a new window]
|
FIGURE 3 Macroscopic currents recorded from wild-type or mutant BKCa channels in excised inside-out membrane patches. cDNAs encoding wild-type (column A), the E912(884)A mutant (column B), the D916(888)A mutant (column C), the D923(895)A mutant (column D), or the E912A/D923A double mutant (column E) forms of the BKCa channel -subunit were transiently expressed in HEK 293 cells as described in Materials and Methods. Current families were recorded in response to increasing concentrations of cytoplasmic Ca2+, which are denoted on the left-hand side. The horizontal dash preceding each family of current tracings indicates the zero current level. For zero Ca2+ conditions (i.e., 2 mM EGTA alone), voltage-clamp steps from –30 to 240 mV, and were given from a holding potential of 0 mV. For 1 and 4 µM free Ca2+ concentrations, steps ranged from –90 to 180 mV and the holding potential was –60 mV. For Ca2+ concentrations between 10 and 1000 µM, voltage steps ranged from –180 to 90 mV, using a holding potential of –120 mV. For all three protocols, the increment between consecutive voltage-clamp steps was 10 mV. The voltage-clamp protocols used experimentally are depicted beneath their respective rows of current tracings. Current records are shown as the average of two independent current-voltage families recorded sequentially at each free [Ca2+]. The number adjacent to the vertical scale bar in each panel indicates current amplitude in nanoamperes. The timescale indicated by the horizontal bar at the bottom of the figure applies to all current tracings. Note that for a given channel type, displayed currents were recorded from the same membrane patch over the range of indicated free [Ca2+]s; exceptions to this are the current tracings shown for the wild-type and E912A/D923A double mutant channels at 100 µM free Ca2+, which were each recorded from different membrane patches. In a few cases, the vertical scaling of current tracings was adjusted to maintain a uniform size for displayed current families; such adjustments reflect differences in current amplitudes observed for wild-type and mutant channels using a particular voltage-clamp protocol and free [Ca2+].
|
|
Plots of conductance-voltage (G-V) relations derived from measurements of the tail current amplitudes for the wild-type and the five mutant BKCa channels described above were fit with single Boltzmann functions, yielding values for the half-maximal voltages of activation (V1/2 value) and the apparent equivalent gating charge (z). A summary of these values (means ± SE) are listed in Table 1 and a plot of the absolute V1/2 values is shown in Fig. 4 A. Due to the modest rightward shifts in activation observed for the E912A and E912A/D923A mutant BKCa channels in the absence of Ca2+ (i.e., 2 mM EGTA alone), we have also expressed the calcium-dependent V1/2 values relative to the 0 Ca value by calculating the differences between V1/2 values in the presence of cytosolic Ca2+ (i.e., 1–1000 µM) and that in the absence of Ca2+ for each channel type (Fig. 4 B). Even when accounting for these modest offsets, it remains evident that the E912(884)A, D923(895)A, and E912A/D923A channel mutants display reduced calcium sensitivity (i.e., a decrease in the ability of increasing concentrations of cytosolic free Ca2+ to evoke leftward shifts of G-V curves along the voltage axis) compared to the wild-type BKCa channel.
View larger version (25K):
[in this window]
[in a new window]
|
FIGURE 4 Mutations predicted to reduce Ca2+ sensitivity alter the half-maximal voltages of BKCa current activation (V1/2 values) in the presence of cytoplasmic free Ca2+. In panel A, V1/2 values for both wild-type and mutant channels were derived from single Boltzmann functions fit to normalized conductance-voltage plots calculated from individual current recordings at each [Ca2+]. In panel B, V1/2 values for each channel type have been "normalized" by calculating the difference between the average V1/2 value obtained in nominally free Ca2+ (A) and the average V1/2 values obtained in the presence of increasing concentrations of cytosolic Ca2+ (e.g., V1/2, 1 µM Ca – V1/2, 0 Ca). Data are presented as means ± SE. The number of membrane patches used to calculate the mean data points at the indicated concentrations of Ca2+ for the wild-type and mutant channels varied as follows: wild-type, five to eight patches; E912A, four to 10 patches; D916A, four to five patches; D923A, four to 10 patches; E912A/D923A double mutant, four to 12 patches.
|
|
Given that the activation and deactivation kinetics of macroscopic BKCa channel currents are Ca2+ dependent (10,12), the observed changes in the activation/deactivation time courses of mutant BKCa channel currents in the presence of cytosolic Ca2+ (see Fig. 3) would be consistent with a reduced calcium sensitivity of these channels. To examine these phenomena in greater detail, the time courses of current activation and deactivation were quantified over broad ranges of membrane voltage and cytosolic [Ca2+]. Fig. 5 A shows representative tracings of deactivating tail currents recorded from wild-type and mutant BKCa channels in the presence of 4 and 100 µM concentrations of Ca2+. Semilogarithmic plots of the time constants derived from single exponential fits of the rising and decaying phases of currents recorded in the presence of Ca2+ concentrations ranging from 0 to 400 µM are shown in Fig. 5, B and C, respectively. In the presence of nominally free Ca2+ (i.e., 2 mM EGTA), no major differences in either activation or deactivation time constants were observed between wild-type and mutant channels. However, over the range of 4–400 µM free Ca2+, it is evident that the E912A and D923A mutants and E912A/D923A double mutant display slower activation kinetics compared to the wild-type channel. Similarly, these same mutants also undergo somewhat faster current deactivation; however, this latter effect does not appear to be as robust as that observed for activation. This observation is thus consistent with recent data from Lingle and co-workers (39) suggesting that the calcium bowl domain may regulate Ca2+-dependent activation, whereas the molecular process underlying current deactivation may be more dependent upon the second high-affinity Ca2+ sensor associated with residues D362/D367 in the RCK domain.
View larger version (31K):
[in this window]
[in a new window]
|
FIGURE 5 Mutations that decrease the Ca2+ sensitivity of BKCa channel gating also reduce the time constants of BKCa current activation and deactivation. Panel A shows representative tail currents recorded from wild-type and mutant BKCa channels at low and high Ca2+ concentrations. In the presence of 4 µM Ca2+, channels were first activated by a step pulse to +160 mV, and tail currents were then recorded at step potentials ranging from –100 to +50 mV, delivered in 10-mV increments. For recordings in the presence of 100 µM Ca2+, a step pulse to +80 mV was used to activate channels, followed by step pulses to tail potentials ranging from –160 to –10 mV. The horizontal dash preceding each family of current recordings indicates the zero current level, and the vertical and horizontal scale bars at the bottom of the panel apply to all tracings. Panel B shows semilogarithmic plots of activation time constants versus membrane potential for macroscopic currents recorded from wild-type and individual mutant channels in the presence of 0, 4, 10, 100, and 400 µM free Ca2+ concentrations. Semilogarithmic plots of the time constants of current deactivation versus membrane potential are displayed in panel C at the same free [Ca2+]s. For panels B and C, symbols denoting the type of BKCa channel are as follows: •, wild-type;  , E912(884)A mutant;  , D923(895)A;  , E912A/D923A double mutant. The activation time constants shown in panel B were derived from single exponential fits to the rising phases of current traces similar to those displayed in Fig. 3. The plotted time constants of current deactivation represent single exponential fits to the decay phases of tail current recordings similar to those shown in panel A. Values represent the means ± SE calculated from four to seven individual membrane patches for each channel type. Solid lines through the data represent nonlinear regression fits, according to the exponential equations: Activation Time Constant = A x e (–qFVm/RT); Deactivation Time Constant = A x e (qFVm/RT), where A = the -value at 0 mV, q = slope of the function. F, Vm, R, and T have the same values as defined earlier. For fits of the activation and deactivation time constants, the derived values of q ranged from 0.26 to 0.38 and 0.30 to 0.41, respectively.
|
|
The effect of mutations on the high-affinity Ca2+-binding sites can be isolated in the presence of elevated cytosolic Mg2+
In addition to channel opening produced via the high affinity Ca2+-binding sites, BKCa gating may also be activated by a low-affinity, divalent cation binding site sensitive to millimolar concentrations of either Mg2+ or Ca2+ ions (35). Subsequently, this gating mechanism has been characterized more extensively (33,40), leading to the identification of the channel's "RCK domain" (17) as this lower affinity Mg2+/Ca2+-binding site (34). To examine the selective activation of BKCa channel gating by low concentrations of cytosolic Ca2+, we took advantage of a strategy in which Ca2+ ion binding to the low-affinity site is effectively competed by high concentrations (e.g., 
10 mM) of cytosolic free Mg2+ ions (3,33,34). Upon saturation of this low-affinity binding site by Mg2+ ions, cytosolic free Ca2+ at concentrations up to 100 µM would be expected to influence channel gating by interacting primarily with the high-affinity sites associated with the calcium bowl and residues D362/D367 (38). Experimentally, families of currents were recorded both in the presence (e.g., 100 µM) and absence (e.g., 2 mM EGTA) of cytosolic Ca2+, under conditions in which the free Mg2+ concentration in the bath was increased from 1 to 10 to 80 mM. Normalized G-V relations were calculated from tail current measurements and V1/2 values were derived from single Boltzmann fits of the data. The degree of leftward shift in the observed G-V relations (e.g., 
V1/2) induced by 100 µM Ca2+ was then utilized as an index of the calcium sensitivity of channel activation due to the high-affinity binding sites, as recently described (3). In the presence of 1 mM Mg2+ (i.e., standard recording conditions), the observed V1/2 value should reflect the actions of 100 µM Ca2+ on both the high-affinity and low-affinity sites. In the presence of 80 mM Mg2+, the degree of leftward shift induced by 100 µM Ca2+ should be due primarily to actions at the high-affinity sites, as Ca2+ binding to the low affinity is prevented by an 800-fold excess of Mg2+. In the presence of 10 mM Mg2+, one may expect an "incomplete block" of Ca2+ binding at the low-affinity site by Mg2+, given that these two cations bind with similar affinities (i.e., in the millimolar range), to this site (33,40).
Fig. 6, A and B, show conductance-voltage (G-V) relations for both the wild-type BKCa channel and the E912A/D923A double mutant in the presence and absence of cytosolic Ca2+, under conditions of increasing concentrations of free Mg2+ ions. The observed shifts in the G-V relations induced by Ca2+ and/or Mg2+ in the wild-type channel are qualitatively consistent with those reported earlier (3,33–35,40). Although increasing Mg2+ was found to produce a similar shift in the activation of the E912A/D923A mutant channel in the absence of Ca2+, little or no shift was evident in the presence of 100 µM Ca2+ (Fig. 6 B). Such an effect has not been previously reported for various other BKCa channel mutations. One possibility that could account for this interesting observation is that the E912A/D923A mutations may reduce the divalent ion binding selectivity of the calcium bowl's high-affinity site, leading to increased displacement of bound Ca2+ by Mg2+ and a progressive reduction in Ca2+-dependent channel activation. In support of this possibility, it has been reported that certain mutations within EF hand motifs are capable of lowering the Ca2+/Mg2+ selectivity ratio (13,16,19).
Fig. 6 C plots the Ca2+-induced V1/2 values observed in the presence of 1, 10, and 80 mM concentrations of cytosolic Mg2+. For currents recorded in the presence of 80 mM Mg2+, we observed a V1/2 of 153.7 ± 4.7 mV for the wild-type channel in the presence of 100 µM Ca2+, compared with only 68.5 ± 1.2 mV for the E912A/D923A double mutant.
Mechanistically, Cui and Aldrich (9) have recently shown that the free-energy contributions provided by membrane voltage (GV) and the binding of cytosolic free calcium (GCa) to BKCa channel activation are additive. As a consequence, the observed leftward shift in G-V relations (i.e., V1/2 value) produced by an increase in the cytosolic [Ca2+] should be due primarily to the change in the contribution of calcium binding to the free energy of channel activation (i.e., GCa = GCa at low calcium – GCa at high calcium). Given that GCa = zeV1/2(Ca), it then follows that GCa can be described by the observed calcium-induced leftward shift in the G-V relations, according to the relationship GCa = (zV1/2 at low calcium – zV1/2 at high calcium)eN, where z = the equivalent gating charge derived from fits of the Boltzmann equation to normalized G-V relations for the wild-type and mutant BKCa channels, e is the elementary charge, and N is Avogadro's number.
Fig. 7 shows a plot of the calculated changes in the free energy of activation due to Ca2+ binding after an increase in the cytosolic [Ca2+] from 0 to 100 µM (e.g., –G(5 nM Ca–100 µM Ca)). For the wild-type BKCa channel, raising the cytosolic [Mg2+] from 1 to 10 to 80 mM produced a modest, but insignificant decrease in the calculated change in the free energy of channel opening due to Ca2+ binding. Consistent with the findings of Cui, Cox and co-workers (3,33,34), this observation suggests that much of the free-energy change induced by 100 µM cytosolic Ca2+ can be ascribed to binding at the two identified high-affinity sites, as preventing Ca2+ binding to the low-affinity site by a large excess of Mg2+ ions has only very modest effects. However, in the case of our E912A/D923A mutant channel, these same calculated calcium-dependent free-energy changes were found to be significantly less at 1, 10, and 80 mM concentrations of Mg2+. Mechanistically, it has been assumed in this analysis that elevated Mg2+ acts principally to prevent Ca2+-dependent channel activation via the low-affinity divalent ion binding site (i.e., RCK domain), without grossly affecting Ca2+ actions via the two high-affinity sites. Although this appears to be largely true in the wild-type channel, the situation may be altered in our E912A/D923A mutant channel, such that high Mg2+ may possibly influence the selectivity and/or function of the high-affinity Ca2+-binding sites (see above). Recently, Cui and co-workers have reported that elevated Mg2+ may interfere with Ca2+-dependent channel activation via the high-affinity sites (33) and such effects may account for the modest decline in the calcium-dependent free-energy changes observed in the presence of elevated Mg2+ (Fig. 7) (3). If such an effect were enhanced at the level of the calcium bowl in the double mutant channel, then our calculations may in fact overestimate the impact of the E912A/D923A mutations on the free-energy change induced by 100 µM Ca2+.
The data described thus far indicate that the calcium bowl mutations E912(884)A and D923(895)A are able to affect the Ca2+-dependent activation of the holo-channel in a synergistic manner. As these residues are predicted to act physically as Ca2+-coordinating ligands, based on our homology model (see Fig. 2 B), it may be also anticipated that these same mutations would reduce the direct binding of Ca2+ ions to this domain. To examine this prediction experimentally, we first generated cDNA constructs encoding a 148-amino-acid fragment from both the wild-type and mutant BKCa channels that encompassed the channel's calcium bowl region. These fragments were expressed in bacteria as 6x His fusion proteins and then purified by Ni2+ NTA chromatography. Purified proteins were resolved by SDS-PAGE, electrotransferred to nitrocellulose membrane, and then incubated in the presence of radioactive 45Ca to examine their Ca2+-binding capacity. We (5) and others (2,4) have demonstrated that such fusion proteins are capable of directly binding 45Ca using a standard overlay assay.
Mutations that reduce the calcium sensitivity of BKCa channel gating also decrease direct Ca2+ binding to the calcium bowl domain
Fig. 8 A shows a blot of 45Ca binding to purified, bacterially expressed protein fragments derived from wild-type and mutant BKCa channel -subunits. Compared to the wild-type fragment, it is evident that 45Ca binding to the fusion proteins derived from the E912(884)A, D923(895)A, and E912A/D923A mutant channels is decreased, whereas binding to the D916(888)A mutant does not appear to be altered. 45Ca binding to purified bovine brain calmodulin (CaM), shown in the far left-hand lane, served as a positive control in our experiments. To compare more precisely the levels of 45Ca binding, we subsequently stained the purified proteins on the nitrocellulose membrane with amino black (Fig. 8 B), and then quantified the actual amount of protein in each lane using image analysis software. The degree of 45Ca binding to each full-length, bacterially expressed protein (i.e., 21 kDa species) was then corrected for the actual amount of protein present on the membrane. For comparison purposes, corrected 45Ca binding to each individual mutant protein has been expressed as a percentage of the corrected 45Ca binding to the wild-type fragment (Fig. 8 C). Based on these analyses, it is evident that the E912(884)A and D923(895)A mutations independently decrease direct 45Ca binding to this high-affinity site, whereas combining these two mutations (i.e., E912A/D923A double mutant) produces a synergistic decrease (>50%) in bound 45Ca.
View larger version (25K):
[in this window]
[in a new window]
|
FIGURE 8 Mutations that decrease Ca2+-dependent BKCa channel gating also reduce direct 45Ca binding to the calcium bowl domain. A 148-amino-acid fragment (861NSPV.. to ..CRVA1009) that included the entire C-terminal calcium bowl region was isolated by polymerase chain reaction from the cDNA encoding the wild-type and mutant BKCa channel -subunits. Individual fragments were expressed as 6x His fusion proteins in bacteria, and similar amounts of purified proteins (15–20 µg) were then resolved by SDS-PAGE and transferred to nitrocellulose membrane. Panel A is an autoradiogram showing 45Ca binding to the wild-type and mutant BKCa channel fusion proteins, along with calmodulin (CaM) as a positive control. In addition to the four alanine-substituted channels described in Fig. 2, we examined two additional BKCa channel mutants, 'DKNDE' and DQDDD/NQNNN. The modified sequences of these two mutants are described below; amino acid substitutions relative to the wild-type channel have been underlined. Panel B shows amido black staining of the nitrocellulose membrane containing the purified wild-type and mutant fusion proteins utilized for 45Ca binding in panel A. The wild-type protein migrates with a relative molecular weight of 21 kDa; modest differences in the electrophoretic migration of fusion proteins derived from mutant channels reflect the removal or addition of acidic residues. The lower molecular weight proteins observed in several lanes represent proteolytic fragments of the full-length fusion proteins. In panels A and B, the positions of molecular weight markers (size in kDa) are indicated by the dashes and numbers on the left-hand side of the panels. (C) Direct 45Ca binding to individual fusion proteins was quantified using a PhosphorImager (Molecular Dynamics, Piscataway, NJ), and then corrected for the amount of protein detected on the same nitrocellulose membrane by amido black staining. 45Ca binding to mutant proteins is expressed relative to wild-type binding (100%); values represent the means ± SE derived from three individual experiments. Wild-type sequence, ELVNDTNVQFLDQDDD; DKNDE mutant, DKNNDTNVDFLEQDDD; DQDDD/NQNNN mutant, ELVNDTNVQFLNQNNN.
|
|
Fig. 8 A also shows the effects of two additional mutations we created within the calcium bowl domain (e.g., DQDDD/NQNNN and 'DKNDE') on direct 45Ca binding. Several independent reports have shown that neutralization/deletion of residues within the Asp-rich sequence of the calcium bowl decreases the apparent calcium sensitivity of BKCa channel activation (3,25,31). Using a bacterially expressed 240 amino acid C-terminal fragment from the BKCa -subunit of Drosophila melanogastor (dSlo), Bian et al. have reported that neutralization of five Asp residues (i.e., DQDDDDD
DQNNNNN) within the calcium bowl significantly decreased direct 45Ca binding to this fragment, but did not have a major effect on the activation of the channel by cytosolic Ca2+ (4). More recently, Bao and colleagues have reported that alanine substitution of acidic residues within the calcium bowl region (i.e., DQDDDDDDQDADDD or DQDDDAD) has significant effects on both Ca2+-dependent channel activation and 45Ca binding to a bacterially expressed GST-fusion protein encompassing this region (2). G-V curves for the DQDDD/NQNNN mutant channel displayed much smaller leftward shifts at 1 and 10 µM calcium compared to the wild-type channel (data not shown), and appeared similar to the G-V curves described for the 5D/5N mutant channel (31,38). Our observation that the DQDDD/NQNNN series of mutations strongly inhibited direct 45Ca binding is also in agreement with the inhibitory effects observed by Bao and colleagues after alanine replacement of residues within this sequence (2).
Authentic EF hands display considerable variation in their primary sequences (13,23), and this variability may account for observed differences in their Ca2+-binding affinities (i.e., KD values), which may range from 0.1 to 1 µM. We thus postulated that it may be possible to "improve" the Ca2+-binding properties of the calcium bowl by substituting amino acids found more typically within EF hand motifs. For example, in the ß-isoform of parvalbumin, replacing neutral residues at the +Z and –Y positions within the EF hand loop sequence by aspartic acids was observed to enhance the Ca2+-binding affinity of the mutant protein (16). Based on such findings, we constructed the 'DKNDE' BKCa channel mutant (described in the legend for Fig. 8) by replacing amino acids in key positions with those found in calmodulin (i.e., refer to EF hand sequence displayed in Fig. 1). In contrast to the results reported for parvalbumin, the 'DKNDE' mutant channel did not display greater 45Ca binding compared with the fusion protein derived from the wild-type channel (Fig. 8 C). Similarly, we observed that a mutant BKCa channel containing the 'DKNDE' series of substitutions did not display enhanced activation in the presence of either 1 or 10 µM free Ca2+ (data not shown).
Collectively, the data described in our study provide a number of additional insights regarding the contribution of calcium bowl residues to the Ca2+ dependence of BKCa channel activation. Specifically, we have: 1), firmly identified the contributions of E912(884) and D923(895) to calcium bowl function by using a broad range of free Ca2+ concentrations to reveal the effects of these individual point mutations on channel activation; 2), demonstrated that mutations decreasing the Ca2+ sensitivity of channel gating also reduce the time constants of channel activation/deactivation, as would be predicted from the noted Ca2+ dependence of current kinetics; 3), observed that the E912(884)A and D923(895)A mutations act in a synergistic manner to decrease both the Ca2+ sensitivity of BKCa channel gating and direct 45Ca binding to the calcium bowl domain, which is consistent with the behavior of multivalent Ca2+ ion binding sites; and 4), described a novel change in the activation of the E912A/D923A double mutant channel in the presence of Ca2+ and elevated Mg2+, which may reflect an interesting effect of these mutations to influence the action of Mg2+ ions on the channel's high-affinity Ca2+-binding sites.
|
DISCUSSION
|
|---|
Over the past several years, considerable evidence has highlighted the contribution of the C-terminal calcium bowl domain to the Ca2+ sensitivity of BKCa channel gating (22). Despite the recognized functional importance of this region, little is known about its underlying structural features, in particular, those that may contribute to its putative high-affinity Ca2+-binding site. Several studies have clearly demonstrated that a series of five Asp residues is critically involved in the function of this domain (2,4,31,32,38), however, it remains unclear exactly how these acidic amino acids contribute to Ca2+-dependent channel activation and Ca2+ ion binding. In an earlier study (5), we had suggested that an amino acid sequence within the calcium bowl may resemble an EF hand motif, based on the positions of potential Ca2+-coordinating ligands (i.e., residues with oxygen-containing side chains) and the similarity of residues to those existing within authentic EF hand structures (13,23). In this study, we have pursued this possibility by employing a homology modeling strategy to generate a predicted structure of this region, using the atomic coordinates of a Ca2+-bound EF hand from human calmodulin as a structural template (see Fig. 2). As anticipated, the resulting three-dimensional model reveals individual acidic residues in this region of the BKCa -subunit that may potentially act as Ca2+-coordinating ligands, based on the spatial orientation of oxygen atoms in their side-chain groups. To evaluate the predictions arising from this model, we have utilized a multifaceted strategy involving site-directed mutagenesis, electrophysiologic measurements of Ca2+-dependent channel activation and the binding of radioactive 45Ca to the calcium bowl expressed as a bacterial fusion protein.
Evaluation of model predictions and experimental data
Electrophysiological recordings of wild-type and mutant BKCa channels in excised membrane patches of HEK 293 cells revealed that the individual E912(884)A and D923(895)A mutations decreased the Ca2+ sensitivity of steady-state BKCa channel gating, as judged by the magnitude of Ca2+-induced leftward shifts in the individual G-V curves (Figs. 3 and 4). In contrast, mutant channels bearing either N918(890)A or Q920(892)A substitutions did not display changes in their Ca2+ sensitivity under the same recording conditions. Although we could not evaluate the N915(887)A mutant, due to poor expression, it was recently reported that this same mutation did not affect the Ca2+ sensitivity of another murine BKCa channel clone expressed in Xenopus oocytes (2). Taken together, these results suggest that the side chains of residues N915(887), N918(890), and Q920(892) do not contribute to the functional Ca2+ sensitivity of BKCa channel activation via the calcium bowl domain. Such observations may reflect both the lack of net charge of these side-chain groups and their predicted orientation away from the bound Ca2+ ion in our model (Fig. 2 B).
Several of the above findings agree closely with the recent results of Bao and colleagues, who utilized alanine scanning mutagenesis to "map" functionally important residues within the calcium bowl region (2). In their study, they noted the reduced Ca2+-dependent activation of the D923(895)A mutant, which we had originally described (5), and the lack of effect of alanine substitutions at residues N918(890) and Q920(892). However, our results further indicate that E912(884) influences the Ca2+ sensitivity of BKCa channel activation at cytosolic [Ca2+]s below 10 µM (refer to Fig. 4); as Bao and colleagues only examined this mutant in the presence of 0 and 10 µM cytosolic Ca2+, it is unlikely that the contribution of E912(884) toward Ca2+-dependent channel activation would have been detected under their experimental conditions.
Combining the E912(884)A and D923(895)A mutations produced a more noticeable reduction in the Ca2+-dependent changes in V1/2 values, as well as a slowing of macroscopic current activation and an acceleration of current deactivation (Figs. 3 and 5). As the Ca2+ dependence of both these kinetic effects is well recognized (10,12), the altered activity of the E912A/D923A double mutant appears to be consistent with a functional decrease in the channel's Ca2+ responsiveness. Structural studies have revealed that the selective coordination of single Ca2+ ions within both proteins and organic calcium chelators, such as EGTA, occurs via multiple oxygen atoms that are arranged to form rigid binding pockets or cavities (13). One could thus postulate that a combination of mutations, such as E912A/D923A or replacement of the Asp-rich region in the calcium bowl (i.e., 5D5N), leads to a greater decrease in the Ca2+ sensitivity of BKCa channel gating, as a result of these mutations on a multivalent binding site. Although Bao and colleagues noted that the Ca2+ sensitivity of BKCa channel gating/energetics was strongly reduced by several of their mutations (e.g., D898A and D900A), the functional consequences of combining such mutations on Ca2+-dependent channel gating over a broad range of Ca2+ concentrations, as well as the kinetics of current activation/deactivation, were not examined.
Using an independent assay, we further observed that the E912(884)A and D923(895)A mutations decreased the amount of radioactive calcium (45Ca) directly bound to a 148-amino-acid fragment of the BKCa channel's C-terminus containing the calcium bowl domain expressed as a bacterial fusion protein (see Fig. 8). Consistent with our electrophysiological data, the E912A/D923A double mutation decreased 45Ca binding more than either substitution alone. In their study, Bao and colleagues reported that E912(884)A produced a modest, but insignificant decrease in 45Ca binding to their bacterially expressed, calcium bowl fusion protein, but they did not examine 45Ca binding to the D923(895)A mutant. Similar to our findings, this group also observed that combining mutations that individually reduced the channel's functional Ca2+ sensitivity (e.g., D926(898) and D928(900)) resulted in a larger decrease in 45Ca binding compared to either mutation alone.
In a canonical EF hand, as found in calmodulin or troponin C, residue D923(895) would function as a bidentate Ca2+-binding ligand, occupying the –Z position in the Ca2+ ion coordination loop. In our model, the side chain of D923(895) is correctly oriented to participate in Ca2+ binding, and replacement of this single residue was observed to reduce the channel's apparent Ca2+ sensitivity. Alanine substitution of E912(884), which would occupy the X position within an EF hand binding loop, was also found to reduce Ca2+-dependent channel activation, but in this case, the effect occurred primarily at low concentrations (e.g., <10 µM) of cytosolic Ca2+. In calmodulin and troponin C, replacement of the acidic residue in this X position decreases both Ca2+ binding and protein function (14,28). In contrast to the predictions of our model, alanine replacement of D916(888) did not alter either the channel's apparent Ca2+ sensitivity, or direct 45Ca binding to a bacterial fusion protein containing this same mutation (Figs. 4 and 8).
Parallels between calcium-sensitive BKCa channel gating and EF hands
Two properties of EF hands that may be of particular relevance to the Ca2+ sensitivity of BKCa channel gating are: 1), the selectivity of their metal ion binding sites, and 2), their ability to undergo state-dependent changes in Ca2+-binding affinity. EF hands are known to bind divalent cations with exquisite selectivity, displaying a typical activation sequence of Ca2+ > Sr2+ » Cd2+, Mg2+, Ba2+ (13). Using a series of mutations capable of disabling each of the BKCa channel's putative high- and low-affinity cation binding sites, Lingle and co-workers (39) have now demonstrated that these three sites can be functionally distinguished based on their selective activation by a series of divalent cations. In particular, it was observed that channel activation via the calcium bowl-associated high-affinity site could be mediated by low concentrations (e.g., 10 µM) of either Ca2+ or Sr2+, but not other divalent metals with smaller atomic radii, such as Mn2+, Cd2+, Co2+, or Ni2+. It is noteworthy that this activation profile of the calcium bowl domain appears to mirror the divalent cation selectivity displayed by a number of EF hand structures. These same investigators also reported that channel activation via the second high-affinity site (D362/D367), located within the channel's RCK domain (17), could be effectively mediated by Cd2+, along with Ca2+ and Sr2+. Accommodation of a smaller Cd2+ ion may suggest that this second site differs in its metal ion coordination geometry compared to the calcium bowl site (13).
Extensive kinetic analyses have indicated that the Ca2+-dependent activation of BKCa channels can be readily explained by an allosteric gating mechanism (22), in which the open state of the channel necessarily displays a higher Ca2+ ion binding affinity than that of the closed state. In this context, it may be revealing that EF hand-containing proteins (e.g., calmodulin) also undergo conformation-dependent increases in Ca2+-binding affinity that typically occur in the presence of a cognate target protein (13). In the holo-BKCa channel, structural rearrangements associated with transition to the open state, or those leading to interaction with another domain, may be sufficient to initiate such an affinity change within a putative EF hand-like binding site. This ability of EF hands to undergo a state-dependent change in their divalent cation binding affinity thus appears to be consistent with a key molecular aspect of the allosteric gating mechanism described for BKCa channels (3,8,29).
An EF hand structural model in relation to other calcium bowl mutations
It may be reasonable to expect that any structural model suggested to underlie the putative high-affinity Ca2+ ion binding site in the calcium bowl domain should also encompass data describing the loss of functional Ca2+ sensitivity and direct 45Ca binding after neutralization of negative charges in the calcium bowl domain, as we (see Fig. 8) and others have demonstrated (2,4). In this context, earlier studies have shown that Ca2+ binding within an EF hand structure can be strongly influenced by neighboring protein surface charges present within the same domain. In the case of the two EF hands of calbindin D9K, neutralizing up to three acidic residues positioned near the two identified Ca2+-binding sites dramatically decreased Ca2+ affinity, even though the side chains of these residues did not directly act as coordinating ligands (18,20). Rather, these negatively charged residues appear to influence Ca2+ binding via long-range electrostatic effects, based on their proximity (8–12 Å) to the bound Ca2+ ions. Similarly, it is conceivable that the Asp-rich sequence in the calcium bowl exerts a similar influence, both within the holo-channel and in bacterially expressed fusion proteins. In our model, the side chains of the first four residues in this region (i.e., from D925(897) to D928(900)) are predicted to lie within 10.5 Å of the bound Ca2+ ion (see Fig. 2 B). If one postulated that the actual high-affinity Ca2+-binding site within the calcium bowl domain consisted of a weak EF hand containing a suboptimal number of coordinating oxygen atoms, then this series of acidic residues would be expected to take on an increased functional importance in maintaining the integrity of the binding site, as a result of their electrostatic influence. However, it is also apparent that this same Asp-rich region does not function simply as a cluster of negative charges, as both the length of these charged side chains, as well as their positions within the series, influence the Ca2+ sensitivity of channel gating and direct 45Ca binding to the calcium bowl (2). For example, replacement of D926(898) or D928(900) by Glu disrupts both Ca2+ sensitivity and 45Ca binding, whereas alanine substitution of D927(899) does not affect Ca2+-dependent activation, but does reduce 45Ca binding. In the holo-channel, D927(899) may be neutralized by pairing with a positively charged side chain, whereas in a bacterial fusion protein, such interactions may be lost. In the case of D926(898) and D928(900), increasing side-chain length may disrupt local conformation within the calcium bowl, leading to decreased Ca2+ binding in both the holo-channel and fusion protein. Alternatively, it is equally conceivable that these aspartic acids act as direct Ca2+-binding ligands, as Bao and colleagues have suggested (2).
In other membrane channels, EF hand motifs have been implicated in the Ca2+ sensitivity of twin-pore K+ channels (11,30) and cardiac sodium channels (37), and the possibility exists that a similar mechanism of Ca2+ sensitivity may be conserved in BKCa channels. However, not all noted EF hand sequences may function in this manner. For example, in voltage-gated L, N, and P/Q-type calcium channels, a C-terminal EF hand motif appears to serve as a "transduction module" that participates in Ca2+-induced inactivation via a calmodulin subunit constitutively bound to the channel's pore-forming -subunit (27).
Can we conclude that the calcium bowl region functions as a Ca2+ sensor?
Although we favor the possibility that neutralization of negative charges at positions E912(884) and D923(895), as well as neighboring aspartic acids (2,4
TOP
ABSTRACT
INTRODUCTION
