Barium ions selectively activate BK channels via the Ca²⁺-bowl site

Abstract

Activation of Ca²⁺-dependent BK channels is increased via binding of micromolar Ca²⁺ to two distinct high-affinity sites per BK α-subunit. One site, termed the Ca²⁺ bowl, is embedded within the second RCK domain (RCK2; regulator of conductance for potassium) of each α-subunit, while oxygen-containing residues in the first RCK domain (RCK1) have been linked to a separate Ca²⁺ ligation site. Although both sites are activated by Ca²⁺ and Sr²⁺, Cd²⁺ selectively favors activation via the RCK1 site. Divalent cations of larger ionic radius than Sr²⁺ are thought to be ineffective at activating BK channels. Here we show that Ba²⁺, better known as a blocker of K⁺ channels, activates BK channels and that this effect arises exclusively from binding at the Ca²⁺-bowl site. Compared with previous estimates for Ca²⁺ bowl–mediated activation by Ca²⁺, the affinity of Ba²⁺ to the Ca²⁺ bowl is reduced about fivefold, and coupling of binding to activation is reduced from ∼3.6 for Ca²⁺ to about ∼2.8 for Ba²⁺. These results support the idea that ionic radius is an important determinant of selectivity differences among different divalent cations observed for each Ca²⁺-binding site.

For the Ca²⁺-activated and voltage-dependent BK-type K⁺ channel, two distinct high-affinity Ca²⁺ regulatory sites, one in each of the two lobes of the large cytosolic C terminus of each BK α-subunit, account for essentially all physiological regulation by cytosolic Ca²⁺ (1–4). Thus, in one tetrameric channel, a total of eight Ca²⁺-binding sites contribute to regulation by Ca²⁺. Each of the two lobes in an α-subunit defines a domain that shares sequence (5, 6) and structural homology (7, 8) with bacterial K⁺ channel/transporter cytosolic domains (6, 9–11). Because such domains regulate conductance to potassium (of bacterial channel/transporters), they are termed RCK domains (12). Recent structures of the BK gating rings (7, 8, 13) confirm the expectation that the BK RCK domains, like the bacterial MthK (6, 9, 11) and TvoK (14) channels, form an octameric assembly. An important difference is that the BK-channel gating ring is formed from a set of four dimers of asymmetric RCK domains, and the bacterial channels use four symmetric dimers.

The first functionally defined high-affinity Ca²⁺ site on BK channels, associated with the RCK2 domain, corresponds to the so-called Ca²⁺ bowl (15), a segment in the distal C terminus that is highly enriched in negative charges. Biochemical (3, 16, 17) and structural (7, 17) information support the view that the Ca²⁺-bowl site participates in Ca²⁺ binding. A structure (resolved to 3.0 Å) based on a single monomer of the cytosolic domain of a single human BK-channel subunit has identified a density, presumably Ca²⁺, coordinated in part by residues D895 and D897 of the Ca²⁺ bowl (7). Both D895 and D897 have been shown by mutational work to play an important role in defining Ca²⁺ sensitivity of the Ca²⁺ bowl (3). The second high-affinity Ca²⁺ site was defined based on the ability of mutations within the RCK1 domain, particularly D367 (1) and later E535 (18), to abolish the component of Ca²⁺ regulation not affected by the Ca²⁺ bowl (1). At present, there is no direct evidence that any residues in RCK1 participate in Ca²⁺ binding. However, guided by a Ca²⁺-free BK gating ring structure (8) in which side chains of D367 and E535 may be suitably positioned for Ca²⁺ coordination, a structural model of Ca²⁺ coordination by these residues has been proposed (18).

A powerful tool for distinguishing properties of different Ca²⁺-binding sites is definition of the divalent cation selectivity of a given site. This idea motivated early studies seeking to understand divalent cation regulation of BK channels before molecular knowledge about different binding sites (19–23). In principle, definition of the binding constants and allosteric coupling constants for activation by a series of divalent cations acting at one or the other Ca²⁺ regulatory site may, when related to structural information, provides insight into state-dependent coordination of divalent cation–binding critical for ligand-dependent allosteric activation. At present, our knowledge about the structural features of the Ca²⁺-binding sites in BK channels precludes such an evaluation. However, examination of the ability of a limited series of divalent cations to activate BK channels either through the mutationally defined RCK1 site or the Ca²⁺ -bowl site have provided some of the stronger evidence that there are two distinct high-affinity Ca²⁺ regulatory sites (4). Of a limited set of divalent cations (see Table S1), only Ca²⁺ and Sr²⁺ were effective at the Ca²⁺ bowl; and Ca²⁺, Cd²⁺, and Sr²⁺ acted through the RCK1 domain. Divalent cations of smaller ionic radius, such as Mn²⁺, Co²⁺, Mg²⁺, and Ni²⁺, were ineffective or only weakly effective at either of the higher affinity sites (4). In contrast, all tested divalent cations enhanced BK current activation through a third low-affinity divalent cation regulatory site (1, 22–24).

Here we tested whether Ba²⁺, a divalent cation of ionic radius larger than Sr²⁺, might reveal differences in the two higher affinity binding sites. Ba²⁺ was previously reported as ineffective in activating BK channels (21). However, Ba²⁺ is also a potent blocker of the BK channel pore (25, 26) such that its ability to activate BK channels might be obscured. Our results show that, not only does Ba²⁺ activate BK channels, this effect is mediated entirely by action at the Ca²⁺-bowl site.

Results

Ba²⁺ Activates BK Channels.

The ability of Ba²⁺ to activate BK channels was tested in inside-out patches. Ba²⁺-containing solutions were buffered with EGTA and no added Ca²⁺, with an estimated free Ca²⁺ of less than 10 nM. Patches were initially exposed to 0 Ca²⁺, 0 Ba²⁺. To reveal activation by Ba²⁺, patches were activated by repetitive (0.5 Hz) steps to +120 mV, with intervening intervals at hyperpolarized potentials (Fig. 1). Application of 10-μM Ba²⁺ resulted in a transient increase in current activation (Fig. 1), followed by a reduction of outward current as Ba²⁺ block reaches steady state. Washout of Ba²⁺ resulted in a slow recovery that was not complete even after 40 s (Fig. 1B). The 40-ms command step allowed visualization of time-dependent block in the presence of Ba²⁺ (Fig. 1 A and C). Two aspects of Ba²⁺ block are of note. First, during development of Ba²⁺ block, peak outward current activated by each test step is indistinguishable from the amount of residual current at the end of the previous test step (red lines in Fig. 1C). Thus, no additional block by Ba²⁺ occurs during the 460 ms between activation steps (20 ms at −120 mV, 290 ms at 0 mV, 150 ms at −140 mV), all conditions at which channel Po is low. Second, 10-μM Ba²⁺ reduces the outward current activated at +120 mV to a level about 10% of that activated with 0 Ca²⁺. Thus, although Ba²⁺ can result in activation of BK channels, block by Ba²⁺ precludes standard evaluation of the activating effects of Ba²⁺.

Fig. 1.

Activation of BK channels by Ba²⁺ precedes block by Ba²⁺. (A) BK currents (wt Slo1) were activated in an inside-out patch with the indicated voltage protocol. A 40-ms step to +120 was preceded by 150 ms at −140 mV. A 20-ms repolarizing step to −120 mV preceded a 290-ms interval at 0 mV. The red trace indicates current activated with 0 Ca²⁺ before application of 10-μM Ba²⁺. Ba²⁺ transiently increased activation of BK current, with time-dependent block during each depolarizing step and gradual diminution in peak current until development of steady-state block. (B) The temporal features of the activation, block, and recovery from parts of three applications of 10-μM Ba²⁺ for the patch shown in A are plotted. Recovery from Ba²⁺ block is not fully complete even after more than 40 s. (C) Traces from the application of 10-μM Ba²⁺ in A are plotted sequentially to highlight both the time course and magnitude of activation and block by Ba²⁺. Red trace is the same as in A. Red lines correspond to the mean amplitude of an averaged segment of current at the end of the preceding trace.

Activation by Ba²⁺ Occurs Through the Ca²⁺-Bowl Site.

We next tested the ability of 10-μM Ba²⁺ to activate channels when either the Ca²⁺-bowl or RCK1 sites were mutated. In both cases, the E399A mutation was included to remove the potential contribution of activation of the low-affinity divalent cation site (1, 24). BK channels containing only a functional Ca²⁺ bowl (E399A/D362A/D367A) are readily activated by 10-μM Ba²⁺ (Fig. 2 A and D), with an activation and block time course similar to that seen for wt BK channels. In contrast, when the Ca²⁺ bowl is mutated (5D5N) together with the E399A mutation, 10-μM Ba²⁺ fails to increase BK activation, while only revealing inhibition by Ba²⁺ (Fig. 2 B and E).

Fig. 2.

Ba²⁺ selectively activates BK channels at the Ca²⁺-bowl site, but not the RCK1 site. (A) Residues that mediate a high-affinity Ca²⁺ effect within RCK1 were mutated, D362A/D367A, along with mutation of a residue that influences low-affinity divalent cation regulation (E399A). Using 15-ms test steps to +120 mV, application of 10-μM Ba²⁺ results in initial activation and then inhibition of BK current, similar to the WT channels; 50-μM Ba²⁺ was applied immediately after 10-μM Ba²⁺, resulting in more complete steady-state inhibition. (B) Mutation of Ca²⁺-bowl residues, 5D5N, along with xE399A mutation completely abolished any activating effect of 10-μM Ba²⁺. The dotted lines correspond to 0 current levels. (C) Example traces correspond to currents for WT BK channels before (black), at the peak of response to 10-μM Ba²⁺ (red), and during block by Ba²⁺ (blue). (D) Current traces are as in C, but for the E399A/D362A/D367A construct. (E) Current traces are as in C, but for the E399A/5D5N construct.

Blockade by Ba²⁺ During Depolarization Is Consistent with Previous Observations of Inhibition by Ba²⁺.

If the block by Ba²⁺ observed in Figs. 1 and 2 arises solely from the strongly voltage-dependent Ba²⁺ block of BK channels described in earlier work (19, 25, 27, 28), at negative potentials it might be possible to exclusively define the ability of Ba²⁺ to activate BK channels without the confounding influence of block. Without undertaking a complete evaluation of Ba²⁺ block, we used a few tests to establish the properties of such block. Whereas sequential 40-ms test steps result in near maximal inhibition within about 5–6 test steps (Fig. S1A), with a 15-ms test step 12–15 test steps are required to produce a comparable fractional inhibition (Fig. S1B). If block by Ba²⁺ only occurs during periods of BK-channel activation, the onset of block plotted as a function of cumulative activation time should be identical for either protocol. Irrespective of whether 15- or 40-ms test steps are used, indistinguishable block rates are observed (Fig. S1C). This is consistent with earlier work indicative that the forward rate of block is dependent on BK-channel open probability (29), but also may simply reflect the very steep voltage-dependence noted for Ba²⁺ inhibition (19). For this example, block during the test steps occurs with an effective forward rate of ∼16.6 s⁻¹ at 10-μM Ba²⁺, corresponding to a first-order rate constant of ∼1.7 × 10⁶ M⁻¹ s⁻¹ at +120 mV. Dependent on the valence of Ba²⁺ block, typically considered to be ∼0.65–0.8e, this corresponds to a 0-voltage block rate of 10⁴−10⁵ M⁻¹ s⁻¹, similar to estimates based on single channel analysis (19). Thus, the inhibition observed here exhibits rates and dependence either on voltage or Po consistent with known features of Ba²⁺ block determined from steady-state single channel methods. Importantly, there is no block occurring during the intervals between the test steps to positive potentials.

We also measured the inhibition rate for block of the E399A/D362A/D367A mutant and the E399A/5D5N mutant (Fig. S1D). The onset of inhibition determined from the decay of current as a function of cumulative activation time was consistently slower for the Ca²⁺-bowl mutant (Fig. S1 D and E). This is the result expected if the 5D5N mutant opens to a lower open probability during application of 10-μM Ba²⁺. The difference in effective blocking rate can be explained if the peak Po at +120 mV in the presence of 10-μM Ba²⁺ for wt and D362A/D367A is about threefold higher than the Po at +120 mV for the 5D5N construct. This is consistent with the expectation for the dependence of Ba²⁺ inhibition on channel Po (29).

Ba²⁺ Activation of the Ca²⁺ Bowl Site: Lower Affinity and Reduced Coupling Compared with Ca²⁺.

Because blockade by Ba²⁺ precludes the use of standard conductance-voltage (GV) curves for estimates of the effect of Ba²⁺, we examined the ligand dependence of BK activation by measurements of NPo at relatively negative potentials (30, 31). Such measurements offer two advantages. First, by measuring NPo at −80 mV, block by Ba²⁺ is minimal. Given an effective δ of about 0.6–0.8e (19) for previous estimates of Ba²⁺ block affinity (30, 31), at 10-μM Ba²⁺ the equilibrium between Ba²⁺ blocked channels and unblocked open channels (OB/O) is 0.0015, and even at 100-μM Ba²⁺ is only 0.015. Thus, the direct reduction in Po due to block even by 100-μM Ba²⁺ will be a little over 1% and will not influence estimates of activation by Ba²⁺. Second, because voltage sensors are predominantly in resting conditions at −80 mV (30), the dependence of NPo on [Ba²⁺] can be directly used to make estimates about Ba²⁺ affinity and coupling of Ba²⁺ binding to channel activation (see Materials and Methods).

Inside-out patches were used to make direct measurements of NPo for three constructs, (i) E399A (Fig. 3A), (ii) E399A/D362A/D367A (Fig. 3B), and (iii) E399A/5D5N (Fig. 3C). For all patches, measurements with 300-μM Ca²⁺ were first used to define the maximal NPo that could be elicited, while 0 Ca²⁺ solutions defined minimal NPo. Estimates of NPo for different Ba²⁺ and Ca²⁺ concentrations were made relative to the NPo at 300-μM Ca²⁺. Because we did not undertake a complete titration with Ca²⁺ for each construct, we normalized our NPo estimates to the absolute Po estimated at ∼300 μM Ca²⁺ in the study by Sweet and Cox (31). For each construct, the fold increase in NPo with Ca²⁺ increases from 0 to 300 μM was similar to previous values (31). Estimates of Po obtained in different concentrations of Ba²⁺ (Fig. 3D) show that, for E399A, activation by 200-μM Ba²⁺ is more than hundredfold less effective than activation by 300-μM Ca²⁺. For E399A/D362A/D367A, titration with Ba²⁺ yields estimates of maximal Po that approach those seen with 300-μM Ca²⁺ but with a somewhat shifted half-activation concentration (Fig. 3 B and D). In contrast, for E399A/5D5N, even 200-μM Ba²⁺ elicits only minor increases in Po (Fig. 3 C and D).

Fig. 3.

Activation by Ba²⁺ at the Ca²⁺-bowl site occurs with an approximately fivefold weaker affinity and reduced coupling. Multichannel patches were held at −80 mV with the indicated divalent cation concentrations for E399A (A) and E399A/D362A/D367A (B), and E399A/5D5N (C). In all patches, NPo was initially measured with 300-μM Ca²⁺ to define maximal activation by Ca²⁺ and then over 100 s of activity was measured in 0 Ca²⁺ or solutions with different [Ba²⁺]. From the maximal conductance activated by 300-μM Ca²⁺ from GV curves at positive potentials, estimated numbers of channels are (A) 98, (B) 216, and (C) 231. Following normalization of NPo measured at 300-μM Ca²⁺ to estimates of absolute Po at 300-μM Ca²⁺ (31), the estimated log(Po) for 200-μM Ba²⁺ for the full traces used for the samples in A–C were −3.71 for E399, −3.89 for E399/D362A/D367A, and −5.96 for E399A/5D5N. (D) Black symbols replot Po estimates as a function of [Ca²⁺] from Sweet and Cox (31) for E399N (black circles), E399N/D367A (black diamonds), and E399N/5D5N (black squares). Red symbols are mean Po estimates as a function of [Ba²⁺] (based on normalization to response at 300-μM Ca²⁺; see Materials and Methods) for E399A (red circles), E399A/D362A/D367A (red diamonds), and E99A/5D5N(red squares). Note that symbols for activation of E399A and E399A/D362A/D367A by Ba²⁺ cannot be readily distinguished.

To estimate effectiveness of Ba²⁺ activation at the Ca²⁺-bowl site, NPo estimates were replotted as log[NPo(Ba²⁺)/NPo(0)] for both the E399A construct and for the construct with an intact Ca²⁺ bowl (E399A/D362A/D367A) (Fig. 4A). Values for Ca²⁺ titration of log[NPo/NPo(min)] generated by Sweet and Cox (31) were plotted for comparison. Fitting with Eq. 2 provided estimates of K_c and K_o, the affinities of Ba²⁺ binding to the closed and open conformations with only a functional Ca²⁺ bowl. Resulting values were 16.6- and 6.6-μM for K_c and K_o, respectively, corresponding to a coupling constant, C, of 2.8. This compares to K_c for Ca²⁺ at the Ca²⁺ bowl of 3.1 μM with a coupling constant, C, of 3.6 (31). Thus, whereas Ca²⁺ can increase BK activation via the RCK2 site by ∼168-fold (C⁴), Ba²⁺ can increase activation by ∼61.5-fold with an approximately fivefold decrease in binding affinity. In contrast, comparison of Ca²⁺ and Ba²⁺ titration of log[NPo/NPo(0)] for the RCK1 site (E399A/5D5N) reveals only a weak ability of Ba²⁺ to produce channel activation (Fig. 4B).

Fig. 4.

Activation of the Ca²⁺ bowl by Ba²⁺ is less effective than activation by Ca²⁺. (A) Activation by Ba²⁺ at the Ca²⁺-bowl site (red diamonds; E399A/D362A/D367A) is replotted as log(NPo/NPo_min) and compared with activation by Ca²⁺ [(31); black diamonds]. For comparison, titrations of channels with both Ca²⁺ sites intact are also shown [black circles: Ca²⁺ on E399N (31)]; red circles, Ba²⁺ on E399A. Solid lines are fits of Eq. 2 (see Materials and Methods) to titration of the Ca²⁺-bowl site with best-fit estimates as shown. A fit of the Ba²⁺ dependence of activation of E399A resulted in K_o = 5.9 μM, K_c = 15.6 μM, yielding C∼2.6. (B) Activation by Ba²⁺ at the RCK1 site (red squares; E399A/5D5N) is compared with activation by Ca²⁺ (31). Solid line is fit to Ca²⁺ titration of the RCK1 site (31) with values as given.

From previous measurements (31), occupancy of each Ca²⁺-bowl site results in a 3.6-fold increase in the C–O equilibrium constant, such that activation of a single site is associated with a free energy difference given by ΔΔG_CO = 0.6ln[C] or 0.94 kcal/mol, for a total of ∼3.76 kcal/mol for all four Ca²⁺-bowl sites. For activation by Ba²⁺, the present results indicate ΔΔG_CO= 0.61 kcal/mol per site, a 0.3 kcal/mol difference from binding of Ca²⁺.

BK activation is commonly assessed by examination of effects of activators on GV curves. Because of the strong voltage-dependence of Ba²⁺ block, this approach is unsuitable in the present case. However, to gain a qualitative estimate of the effectiveness of Ba²⁺ as a BK activator for comparison with the results just described, we used a procedure to attempt to minimize block by Ba²⁺, while allowing estimates of the foot of GV curves in the presence of either Ca²⁺ or Ba²⁺ (Fig. S2A). The stimulation sequence was designed to allow definition of a GV curve first in 0 Ca²⁺ and then in a given [Ba²⁺], returning to resting conditions between each test step in Ba²⁺ (Fig. S2A). Using this protocol, for Slo1 wt channels, 10-μM Ba²⁺ resulted in an increase in conductance that was comparable to that produced by 1-μM Ca²⁺ (Fig. S2B). For a construct with no intact Ca²⁺ regulatory sites (Slo1-E399A/D362AD367A/5D5N), no activation by Ba²⁺ was detected (Fig. S2C). For an Slo1 construct with only an intact Ca²⁺ bowl (Slo1-E399A/D362A/D367A) (Fig. S2D), Ba²⁺ is clearly more effective than 1-μM Ca²⁺ at producing channel activation. In contrast, for Slo1-E399A/5D5N, Ba²⁺ did not increase activation of conductance (Fig. S2E), and similarly there was no obvious activation by Ba²⁺ when only E399 was intact (Slo1-D362AD367A/5D5N). These tests suggest that the effect of 10-μM Ba²⁺ at the Ca²⁺-bowl site is approximately equivalent to the effect of perhaps 2–4-μM Ca²⁺. The approximately fivefold higher [Ba²⁺] required to produce a GV shift comparable to that in a given [Ca²⁺] is generally consistent with the direct estimates of Ba²⁺-binding affinity obtained from NPo measurements (Fig. 4).

Discussion

The present results establish that the extensively studied (19, 25–27, 29, 32–34) K⁺-channel blocker, Ba²⁺, is also an activator of BK channels. That activation of BK channels by Ba²⁺ was overlooked in earlier studies is not surprising, because a long-preferred method for investigation of BK channels is simply to monitor activity of single channels under steady-state conditions (19, 21, 25–27). Typically, this is done at positive voltages to favor BK activation, a condition which in the presence of Ba²⁺ allows block by Ba²⁺ to dominate closing events, thereby obscuring the Ba²⁺ enhancement of activation.

The most important aspect of the present results is that Ba²⁺ uniquely distinguishes between the two high-affinity Ca²⁺-binding sites, producing selective activation of BK channels via the Ca²⁺-bowl site on RCK2. This contrasts nicely with selective action of the RCK1 site by Cd²⁺ (4). Together these results suggest that, for the ions that have so far been examined, the preferred ionic radii of ions able to produce activation via the RCK1 site are smaller than those effective at the Ca²⁺-bowl site (Fig. 5A). However, differences in the chemistry of different divalent may also contribute. As yet, the available structural information for the BK-channel gating ring is insufficient to allow any correlation of ion selectivity with structural features of the Ca²⁺-binding sites. Although a density likely to correspond to Ca²⁺ has been observed in coordination with particular Ca²⁺-bowl residues (7, 13), resolution of structural features in this region is limited. However, it is interesting to consider potential changes that may occur at the Ca²⁺ bowl during gating (Fig. 5B). Atoms in closest contact with the Ca²⁺ ion include the main-chain carbonyls from Q889 and D892 and side-chain carboxylate groups of D895 and D897 (7). Both D895 and D897 were shown by mutagenesis to be critical for Ca²⁺-dependent gating in this region (3). Despite somewhat limited side-chain resolution in the available Ca²⁺-bowl structures, it is perhaps noteworthy that, based on the two available structures, the side chains of both D895 and D897 must undergo a greater than 5-Å displacement (rotation) to interact with Ca²⁺. This raises the possibility that the size of the coordinating divalent cation and the length of the resulting metal–oxygen bonds may be critical for stabilizing the Ca²⁺ bowl in a particular liganded conformation. Divalent cations of smaller radii may be unable to provide the interactions necessary to stabilize the active conformation.

Fig. 5.

Selectivity among BK channel divalent cation–binding sites. (A) For each putative divalent cation site, the ionic radii of divalent cations effective at each site are plotted. Sizes of circles are scaled to reflect estimates of ionic radius. Ba²⁺ at up to 200 μM is ineffective on the E374/E399 site, and higher concentrations have not been tested. (B) Models of the Ca²⁺-bowl region are shown for the Ca²⁺-free structure (3NAF) (8) and for a structure of a gating ring monomer obtained in high Ca²⁺ (3MT5) (7). Structures were overlaid based on the minimal RMS deviation of α-carbons from residues 885–905. Text label identifies coordinating oxygens (spheres) along with the distance in Ångstroms (in parentheses) that the oxygen moves between the Ca²⁺-bound and Ca²⁺-free structures. Yellow sphere corresponds to an ion with a radius of 0.99 Å.

What differences in the Ca²⁺-bowl structure might be expected based on occupancy by Ca²⁺ vs. Ba²⁺? For a set of six PDB structures with bound Ca²⁺ (see SI Text), the average Ca–O distance (for 37 Ca–O bonds) was 2.39 ± 0.19 Å. For a set of five PDB structures with bound Ba²⁺, the average Ba–O distance (for 35 Ba–O bonds) was 3.16 ± 0.38 Å. With the assumption that Ba²⁺ may be coordinated by at least some of the same residues that coordinate Ca²⁺, this suggests that occupancy of the Ca²⁺ bowl by Ba²⁺ would place Ba²⁺ in a position displaced at least 0.7 Å from the D889 and D892 main-chain carbonyls, which in turn would impact on the approach of D895 and D897 side chains. Thus, a Ca²⁺-liganded Ca²⁺ bowl is likely to exhibit a more compact structure than a Ba²⁺-liganded Ca²⁺ bowl, perhaps impacting overall gating ring structures in the two cases. That the liganded Ca²⁺-bowl structure may differ in the two cases provides a potential explanation for the observation that the limiting Po for Ca²⁺ and Ba²⁺ activation mediated via the Ca²⁺ bowl differs. Although at saturating ligand, the Ca²⁺-bowl site may be similarly occupied either by Ba²⁺ or Ca²⁺, the Ba²⁺-liganded Ca²⁺-bowl structure is apparently less effective at producing a gating ring conformation conducive to full channel opening.

Spectroscopic and optical approaches have also been used to examine conformational changes in an isolated full BK gating ring apparatus (35). Whereas Ca²⁺ up to 35 μM produces a quenching of the emission spectra consistent with a Ca²⁺-dependent conformational change in the full gating ring, Ba²⁺ was without effect up to 190 μM, with only modest quenching even at 13 mM Ba²⁺. The inability of Ba²⁺ to produce spectroscopic changes similar to those produced by Ca²⁺ seems difficult to explain within the context of our observations of direct activation of BK channels by μM Ba²⁺. One trivial explanation might be that the activation we observe with Ba²⁺ is mediated by some other contaminant in our solutions. However, this would require that such a contaminant exhibit selectivity between the two Ca²⁺ regulatory sites distinct from all of the other activating divalent cations so far examined. The properties of the block we observe clearly confirm that Ba²⁺ is the principal component of our test solutions. Another possibility, perhaps consistent with the less effective activation by Ba²⁺ at the Ca²⁺ bowl, is that the conformational change associated with Ba²⁺ binding at the Ca²⁺ bowl may not produce the structural changes required for changes in emission spectra. Although we have no explanation for the absence of Ba²⁺-induced spectroscopic changes in isolated gating rings, the conditions of the spectroscopic measurements and the present experiments differ in one important way. Namely, the spectroscopic experiments reflect a gating ring isolated from the pore-gate domain.

Despite the relative simplicity of divalent cations as ligands, multiple factors may influence the ability of a given binding site to distinguish effectively among divalent cations. This would include the properties, position, and number of ion-coordinating residues, access paths to the ion-binding residues, and the physical and chemical properties of the ion itself, including ionic radii and rates of dehydration. At present, the most well-developed analysis of structural determinants of Ca²⁺ binding have been done with the EF-hand Ca²⁺-binding domains, which offer some interesting guides for consideration of other Ca²⁺-binding motifs. Systematic manipulations on both C2 and EF-hand Ca²⁺-binding domains have presented compelling examples of how both site access and coordinating residues can influence not only affinity but also selectivity (36–42). Furthermore, for isolated Ca²⁺-binding domains from calmodulin, NMR studies have also provided interesting guidance regarding the importance of divalent cation size in permitting formation of an activated conformation (43–45). Mg²⁺, which has negligible ability to activate calmodulin, is able to bind with low affinity, but in a fashion that mimics the ligand-free apo state, failing to coordinate with the 12th-position acidic residue, critical for Ca²⁺-dependent activation. Consistent with the basic idea of ligand-dependent allosteric regulation, the existence of these conformations highlights the idea that, for any ligand-binding site, at least four conformations are expected: two closed forms, apo and liganded, and then two open forms, apo and liganded.

For the BK-channel Ca²⁺-binding sites, the use of divalent cations other than Ca²⁺ promises insight into how binding-site occupancy may impact on gating ring structure. Furthermore, the availability of a well-defined allosteric model that allows experimental definition of state-dependent binding constants and allosteric coupling constants provides a critical tool for a potential detailed analysis of the consequences of manipulation of the ligand-binding sites.

Materials and Methods

Oocyte Removal and Culture.

Stage IV Xenopus laevis oocytes were used for expression of all constructs. Mutated Slo1 constructs, 5D5N and D362A/D367A, with and without the E399A mutation were prepared as previously described (1, 46). We compared our results to Ca²⁺ titration curves obtained with constructs in which D367A alone was used to disrupt RCK1 Ca²⁺ sensing and E399N to disrupt the low-affinity binding site (31). Constructs E399A and E399N exhibit comparable reductions in the ability of high Mg²⁺ and Ca²⁺ to shift BK gating (1, 24). Similarly, the D367A mutation alone is comparable to that of simultaneous mutation of D362A/D367A (1).

Electrophysiology.

Currents were recorded in the inside-out configuration (47) using an Axopatch 200 amplifier (Molecular Devices) and the Clampex program from the pClamp software package (Molecular Devices). Recording pipettes used borosilicate capillary tubes (Drummond Microcaps, 100 μL), pulled to diameters of typically 0.5–1 μM, coated with Sylgard (Sylgard 184, Dow Chemical Corp.) and fire-polished. Gigaohm seals were formed in normal frog Ringer (in millimolars, 115 NaCl, 2.5 KCl, 1.8 CaCl₂, and 10 Hepes at pH 7.4) and, after excision, moved into flowing test solutions with solution application as used previously (48, 49). The pipette/extracellular solution was (in millimolars) 140 K-methanesulfonate, 20 KOH, 10 Hepes, and 2 MgCl₂ at pH 7.0. Test solutions bathing the cytoplasmic face of the patch membrane contained (in millimolars) 140 K-methanesulfonate, 20 KOH, 10 Hepes, with pH adjusted to the nominal value with KOH. Ba²⁺-containing solutions were buffered with EGTA and composition of solutions was calculated using EQCALWIN (V. 1.1, Biosoft), assuming stability constants of 8.41 (BaEGTA) and 10.89 (CaEGTA). Assuming contaminant-free Ca²⁺ in water of ∼10-μM, free Ca²⁺ in the buffered Ba²⁺ solutions is less than 10 nM.

Evaluation of PDB Models.

Structural alignment, image rendering, and measurement of atomic distances were performed using University of California, San Francisco Chimera (50).

Estimates of Ba²⁺-Binding Constants.

For estimates of NPo, we used steady-state recordings from inside-out patches held at −80 mV. NPo was defined either by measurements of half-amplitude crossings or from distributions of total current amplitudes. Both procedures yielded similar estimates. For each patch, NPo in the presence of 300-μM Ca²⁺ was first determined, defining maximum channel activation. To estimate Po, NPo estimates in the presence of 300-μM Ca²⁺ were normalized to similar Ca²⁺ concentrations from previous work (31). For all constructs, the fold increase in NPo from 0 Ca²⁺ to 300-μM Ca²⁺ was comparable (see Fig. 3) to that previously reported (31).

Based on the Horrigan-Aldrich model for BK activation (30), at negative potentials with two high-affinity Ca²⁺-binding sites and the assumption that Po is small over the entire range of [Ca²⁺], the dependence of (Po/Po_min) on Ca²⁺ concentration is defined by the following (31):

where K_O1 and K_C1 represent the open- and closed-state affinity of site 1 for Ca²⁺, and K_O2 and K_C2 reflect the site 2 affinities. For each individual site

At [Ca²⁺], which maximally activates a given site (Ca_max), from Eq. 2, the ratio of Po(Ca_max)/Po(0) is defined simply by K_c⁴/K_o⁴, which is equivalent to C⁴, where C is the allosteric constant reflecting coupling of Ca²⁺ binding to a given site to channel activation. The relationships just defined also apply to patches with an indeterminate number of channels because Po(Ca)/Po(0) = NPo(Ca)/NPo(0). The dependence of log[NPo(Ba)/NPo(0)] on [Ba²⁺] was therefore fit with a form of Eq. 2 to define the closed- and open-state binding affinities. The allosteric constant C was defined from the relationship between the binding affinities.