There is controversy over whether Ca2+ binds to the BKCa channel's intracellular domain or its integral-membrane domain and over whether or not mutations that reduce the channel's Ca2+ sensitivity act at the point of Ca2+ coordination. One region in the intracellular domain that has been implicated in Ca2+ sensing is the “Ca2+ bowl”. This region contains many acidic residues, and large Ca2+-bowl mutations eliminate Ca2+ sensing through what appears to be one type of high-affinity Ca2+-binding site. Here, through site-directed mutagenesis we have mapped the residues in the Ca2+ bowl that are most important for Ca2+ sensing. We find acidic residues, D898 and D900, to be essential, and we find them essential as well for Ca2+ binding to a fusion protein that contains a portion of the BKCa channel's intracellular domain. Thus, much of our data supports the conclusion that Ca2+ binds to the BKCa channel's intracellular domain, and they define the Ca2+ bowl's essential Ca2+-sensing motif. Overall, however, we have found that the relationship between mutations that disrupt Ca2+ sensing and those that disrupt Ca2+ binding is not as strong as we had expected, a result that raises the possibility that, when examined by gel-overlay, the Ca2+ bowl may be in a nonnative conformation.
The large-conductance Ca2+-activated K+ channel (BKCa) provides a link between chemical and electrical signaling by opening in response to micromolar concentrations of Ca2+ close to the plasma membrane (Latorre et al., 1989). The primary sequence of the channel's pore-forming (Slo) subunit, however—four of which form a fully functional channel—contains no established Ca2+ binding motifs (Atkinson et al., 1991; Adelman et al., 1992; Butler et al., 1993; Pallanck and Ganetzky, 1994), and so the origin of the BKCa channel's Ca2+ sensitivity has been of considerable interest as it likely involves structurally novel Ca2+-binding sites.
Slo's secondary structure (Fig. 1)
suggests that a large intracellular carboxy-terminal domain unique to this channel has evolved to confer Ca2+ sensitivity upon what is otherwise a voltage-gated K+ channel (Butler et al., 1993; Wei et al., 1994; Schreiber et al., 1999). Indeed, the BKCa channel is considerably voltage sensitive (Barrett et al., 1982; Methfessel and Boheim, 1982), and mutations in its intracellular domain reduce and, in combination, eliminate its Ca2+ sensitivity (Schreiber and Salkoff, 1997; Bian et al., 2001; Bao et al., 2002; Xia et al., 2002). Contrary to this view, however, it has been found that the Slo subunit can be truncated just after its putative S6 helix, eliminating the channel's entire intracellular domain, and the resulting construct, although it expresses poorly, produces channels with near wild-type Ca2+ sensitivity (Piskorowski and Aldrich, 2002). This argues that the BKCa channel's Ca2+-binding sites reside entirely in Slo's integral-membrane domain and therefore that mutations in the intracellular domain that affect Ca2+-dependent channel opening must do so indirectly.
Despite this important result, however, there are still compelling reasons to think that the BKCa channel's intracellular domain might bind Ca2+. First, the crystal structure of the bacterial Ca2+-activated K+ channel, MthK, reveals an intracellular “RCK” domain that is similar to portions of Slo's intracellular domain, and this RCK domain forms the bacterial channel's Ca2+-binding “gating ring” (Jiang et al., 2001, 2002). Second, mutations in and around Slo's RCK domain decrease the BKCa channel's Ca2+ sensitivity (Wei et al., 1994; Shi and Cui, 2001; Zhang et al., 2001; Bao et al., 2002; Shi et al., 2002). And third, several experiments suggest that an acidic region in Slo's intracellular domain termed the “Ca2+ bowl” may form a Ca2+-binding site: Fairly large mutations in the Ca2+ bowl reduce the channel's Ca2+ sensitivity by approximately half (Bian et al., 2001; Bao et al., 2002; Xia et al., 2002). When a portion of Slo that includes the Ca2+ bowl was transferred to the Ca2+-insensitive Slo3 subunit, Ca2+ sensitivity was conferred upon the previously insensitive channel (Schreiber et al., 1999). And peptides composed of portions of Slo that include the Ca2+ bowl bind Ca2+ in gel-overlay assays (Bian et al., 2001; Braun and Sy, 2001), and this binding is inhibited by the mutation of five Ca2+-bowl aspartic acids (Bian et al., 2001).
Thus, there is evidence both for and against the notion that Ca2+ binds to the BKCa channel's intracellular domain, specifically to the Ca2+ bowl, and in general there is a good deal of confusion as to where Ca2+ binds to the BKCa channel and to what degree mutations that affect Ca2+-dependent channel opening are acting at the point of Ca2+ coordination. If, however, the Ca2+ bowl forms a functionally relevant Ca2+-binding site, then we might make two predictions. First, residues in the Ca2+ bowl that coordinate Ca2+ should in general be more sensitive to mutation than those that do not directly contact Ca2+ (Falke et al., 1994). And second, a correlation should exist between Ca2+-bowl mutations that affect Ca2+ sensing and those that affect Ca2+ binding. To test these predictions, and to perhaps identify residues involved in Ca2+ binding, we made a series of point mutations in the mouse Slo (mSlo) Ca2+ bowl and analyzed the effect of each on the energy-change that Ca2+ binding imparts to the mSlo channel's closed-to-open conformational change. Also, we have examined, by gel-overlay, the effect of many of these mutations on Ca2+ binding to a fusion protein that contains a portion of mSlo's COOH-terminal tail. Our results reveal a subregion of the Ca2+ bowl that is critical for both Ca2+ sensing and Ca2+ binding, and they identify two aspartates in this subregion that are essential for both processes. Thus, much of our data suggests that the Ca2+ bowl forms a functionally relevant Ca2+-binding site. Overall, however, we have found that the relationship between mutations that disrupt Ca2+ sensing and those that disrupt Ca2+ binding is not as strong as we had expected, a result that raises the possibility that, when examined by gel-overlay, the Ca2+ bowl may be in a nonnative conformation.
Materials And Methods
Channel Expression and Electrophysiology
All electrophysiology experiments were done with the mslo clone mbr5 (Butler et al., 1993) essentially as described (Bao et al., 2002). In vitro transcription was performed with the mMessage mMachine kit with T3 RNA polymerase (Ambion). To record macroscopic currents ∼0.5–50 ng of cRNA were injected into Xenopus laevis oocytes (stage IV-V) 2–6 d before recording.
All recordings were done in the inside-out patch-clamp configuration (Hamill et al., 1981). Patch pipettes were made of borosilicate glass (VWR micropipettes), and had resistances of 1–2 MΩ in our recording solutions. Their tips were coated with sticky wax (Sticky Wax) and fire polished before use. Data were acquired using Axopatch 200B patch-clamp amplifiers (Axon Instruments, Inc.) and Macintosh-based computer systems that use “Pulse” acquisition software (HEKA Electronik) and the ITC-16 hardware interface (Instrutech Scientific Instruments). Records were digitized at 50 Khz and low pass filtered at 10 KHz. All experiments were performed at room temperature, 22–24°C. Before current records were analyzed and displayed, capacity and leak currents were subtracted using a P/5 leak subtraction protocol with a holding potential of −120 mV and voltage steps opposite in polarity to those in the experimental protocol.
All mutations were made with the QuickChange site-directed-mutagenesis kit (Stratagene), and mutations were identified by sequencing around the point of the mutation.
Recording solutions were composed of the following (in mM): pipette solution, 80 KMeSO3, 60 N-methyl-glucamine-MeSO3, 20 HEPES, 2 KCl, 2 MgCl2, pH 7.20. Internal solution, 80 KMeSO3, 60 N-methyl-glucamine-MeSO3, 20 HEPES, 2 KCl, 1 HEDTA or 1 EGTA, and CaCl2 sufficient to give the appropriate free Ca2+ concentration; pH 7.20). EGTA (Sigma-Aldrich) was used as the Ca2+ buffer for the 0.003 μM [Ca2+] solutions. HEDTA (Sigma-Aldrich) was used as the Ca2+ buffer for solutions containing 0.8 and 10 μM free [Ca2+]. 50 μM (+)-18-crown-6-tetracarboxylic acid (18C6TA) was added to all internal solutions to prevent Ba2+ block at high voltages.
The appropriate amount of total Ca2+ (100 mM CaCl2 standard solution; Orion Research, Inc.) to add to the base internal solution containing 1 mM HEDTA to yield the desired free Ca2+ concentration was calculated using the program Max Chelator (Bers et al., 1994), which was downloaded from (www.stanford.edu/~cpatton/maxc.html), and the proton and Ca2+-binding constants of Bers (supplied with the program) for pH = 7.20, T = 23°C, and ionic strength = 0.15. The ability of 18C6TA to chelate Ca2+, as well as K+ and Ba2+, was also considered in these calculations using the following dissociation constants: Ca2+ 10−8 M (Dietrich, 1985), K+ 3.3 × 10−6 M (Dietrich, 1985), Ba2+ 1.6 × 10−10 M (Diaz et al., 1996). Free [Ca2+] was measured with a Ca-sensitive electrode (Orion Research, Inc.), and the measured value reported. Endogenous [Ca2+] in our internal solution before addition of Ca2+ chelator was estimated from the deviation from linearity of the Ca-sensitive electrode's response at 10 μM added [Ca2+], and was 16–20 μM. Endogenous [Ca2+] was then compensated for when making Ca2+-buffered solutions. 1 mM EGTA was added to the internal solution intended to contain 0.003 μM free Ca2+, and no Ca2+ was added.
During our experiments the solution bathing the cytoplasmic face of the patch was exchanged using a sewer-pipe flow system (DAD 12) purchased from Adams and List Assoc. Ltd.
G-V relations were determined from the amplitude of tail currents measured 200 μs after repolarization to a fixed membrane potential (−80 mV) after voltage steps to the indicated test voltages. Each G-V relation was fitted with a Boltzmann function (G = Gmax/(1 + e−zF(V − V1/2)/RT))) and normalized to the peak of the fit. All curve fitting was done with “Igor Pro” graphing and curve fitting software (WaveMetrics, Inc.) using the Levenberg-Marquardt algorithm to perform nonlinear least squares fits. This software was also used to fit Eq. 2 to the G-V relations of each experiment as described with reference to Fig. 4. The constant parameters used for these fits were as follows: JC(0) = 0.059; JO(0) = 1.020; z = 0.51; q = 0.4.
To determine the statistical significance of the data in Figs. 4–6, Student's t statistic (difference in sample means/standard error of difference of sample means) was calculated according to Eq. 1 below for either ΔV1/2 or ΔΔGCa and compared against the t distribution with degrees of freedom: sum of all n – 4.
Expression and Purification of GST-fusion Proteins
An EcoRI-NotI fragment encoding 207 amino acids from the COOH terminus of the mouse Slo (mbr5) subunit ranging from ASNFHY to GATPEL was subcloned in frame with the glutathione-S-transferase (GST) of the bacterial expression plasmid pGEX-4T-1 to make the GST-mSlo207 expression vector. The QuickChange site-directed mutagenesis kit (Stratagene) and standard subcloning techniques were used to create the GST-mSlo207 mutant expression constructs. After transformation into Escherichia coli strain BL21, a 6 ml LB/amp (100 μg/ml) overnight culture was used to inoculate 200 ml of LB/amp. The culture was grown for 1.5–2 h (A600 = 0.8) at 37°C and then induced with isopropylthio-β-galactoside (IPTG) (0.1 mM) for 4 h. Bacterial cells were collected and frozen. Cells were thawed at room temperature and resuspended in 10 ml of ETN washing buffer (20 mM Tris, pH 8.0; 100 mM NaCl; 1.5 mM EDTA; 0.1% sarcosyl; protease inhibitors). Cells were kept on ice for 15 min. Additional EDTA (final concentration is 5–6 mM) and sarcosyl (final concentration is 1.4%) were added. Cells were sonicated on ice and then spun down at 10,000 rpm for 50 min at 4°C. Supernatants were collected and poured into 50-ml tubes, then divided into two tubes. To each fraction, 10 ml of 10% Triton X-100 was added and gently mixed. One fraction was frozen at −80°C. To the other fraction, 500–700 μl of blocked glutathione agarose beads (Sigma-Aldrich) were added and incubated with rocking for 2 h at 4°C. The beads were then washed three times with 10 ml of cold PBS (centrifuge at 1,000 rpm for 2 min, and rock for 5 min at 4°C in between washes). Beads were stored at 4°C in an equal volume of cold PBS with the addition of 50 μl of 2% NaN2. Protein was eluted from the beads with 1% SDS in PBS, and concentrated with vivaspin ultrafiltration concentrators (Vivascience). Protein concentrations were then measured by Bradford Assay (Bio-Rad Laboratories).
Ca2+ overlay assays were done essentially as described by Braun and Sy (2001). After SDS-PAGE, protein bands—between 10 and 60 μg per lane— were electroblotted onto a nitrocellulose membrane, which was then dried for 30 min at room temperature. The effectiveness of protein transfer was examined by ponceau staining (Sigma-Aldrich). The stain was then removed by washing the blot with PBS, and the blot was washed further at room temperature 4 × 10 min in 30 ml wash buffer (10 mM imidazole-HCl, 70 mM KCl, pH 6.8; this solution was treated with 1 g/l chelex 100 [Bio-Rad Laboratories] to reduce the amount of contaminant Ca2+), after which the membranes were rocked at room temperature in 30-ml wash buffer with 10 μCi/ml added 45CaCl2 (ICN) (between 9 and 12 μM) for 1 h. This was followed by a single 5-min wash in 50% ethanol. After the blot was hung dry overnight, it was exposed to a phosphorscreen for 6–7 h. Bands were detected by the PhosphorImager system of Molecular Dynamics, Inc. The relative amounts of 45Ca2+ binding were quantified using the densitometric software ImageQuant also from Molecular Dynamics, Inc. Due to problems with consistent washing across blots, we were unable to generate Ca2+-binding curves with this assay.
The Ca2+-bowl residues 896–907 where modeled according to the backbone carbon trace of the first Ca2+-binding loop of parvalbumin (PDB entry 2PVB, residues 51–62). Alignment and energy minimization was done with InsightII software (Accelrys, Inc).
The BKCa channel opens in response to both changes in membrane potential and internal Ca2+ concentration ([Ca2+]) such that increases in [Ca2+] shift the BKCa channel's G-V relation leftward along the voltage axis (Fig. 2
A). For a given change in [Ca2+] the magnitude of the shift depends on the three factors: the voltage sensitivity of the channel, the energetics of Ca2+ binding, and the number of binding sites that influence opening (Cox et al., 1997). Recent analyses of mSlo mutants have suggested that three types of Ca2+-binding sites influence opening—one of low affinity that starts to affect G-V position at ∼10 μM [Ca2+] and two of higher affinity, both of which start to affect G-V position at ∼0.1 μM [Ca2+] (Zhang et al., 2001; Bao et al., 2002; Shi et al., 2002; Xia et al., 2002). Given the fourfold symmetry of the channel (Shen et al., 1994), and recent results with hybrid channels (Niu and Magleby, 2002), it seems likely that there are four of each type of binding site.
One type of high-affinity Ca2+-binding site can be functionally disrupted by point mutations in and around the channel's RCK domain (Bao et al., 2002; Xia et al., 2002). Such mutations reduce by approximately half the G-V shift observed in response to micromolar [Ca2+], as expected if these mutations cause the loss of half of the channel's functionally similar, high-affinity Ca2+-binding sites. This is illustrated in Fig. 2 B with the mutant M513I. The second type of high-affinity site can be similarly disrupted by large mutations in the acidic region of the Ca2+ bowl (Schreiber and Salkoff, 1997; Bian et al., 2001; Bao et al., 2002; Xia et al., 2002). This is illustrated in Fig. 2 C with the deletion mutant Δ896–903 (Bao et al., 2002). When mutations at both positions are made together, all high-affinity response is lost (Fig. 2 D) (Bao et al., 2002; Xia et al., 2002). This result is important in the present context, because it indicates that the mSlo channel contains no other high-affinity Ca2+-binding sites. Thus, if we want to study the BKCa channel's Ca2+-bowl–related Ca2+-binding sites in isolation, or more precisely those sites whose functional effects are altered by mutations in the Ca2+ bowl, we may do so by working with a channel that carries the M513I mutation. This is the approach we have taken in this study. Also, to prevent interference from the channel's low-affinity sites we have restricted our analysis to Ca2+ concentrations less than or equal to 10 μM (Zhang et al., 2001; Bao et al., 2002).
An Alanine Scan of the Ca2+ Bowl
Shown in Fig. 1 C is the sequence of the “Ca2+ bowl” as delineated by Schreiber and Salkoff (1997). Residues with negatively charged side-chains at neutral pH are indicated in red; other oxygen-containing side chains are shown in green. As Ca2+-binding sites in proteins are universally formed by oxygen atoms—typically six or seven coordinate the Ca2+ ion, one or two of which are supplied by water (Falke et al., 1994; Nalefski and Falke, 1996; Perisic et al., 1998)—it seems likely that if the Ca2+ bowl forms a Ca2+-binding site, then some of these oxygen-containing side chains are involved. From the Δ896–903 mutant's (plus M513I) complete loss of Ca2+ sensitivity (Fig. 2 D) we know that one or more of the eight amino acids deleted in this mutant (underlined in Fig. 1 C) are critical for Ca2+-bowl function. Seven of these residues contain side-chain oxygens and six of them are acidic. Similar results have been obtained for the smaller deletion mutant Δ899–903 (Bao et al., 2002) and for the substitution mutant 897–901N (Xia et al., 2002). The structures of residues D899, D900, and D901 or some subset of these, therefore, must be important for proper Ca2+ sensing. Effects on Ca2+ sensing have also been reported, however, for mutations at D897 and D898 (Schreiber and Salkoff, 1997), so perhaps these residues are important as well, and clearly, as the great majority have not been tested, there could be many other residues in the Ca2+-bowl whose side-chains are also involved in Ca2+ sensing.
To examine each residue individually, we mutated each oxygen-containing side-chain in the Ca2+ bowl (each red or green residue in Fig. 1 C) to alanine, one at a time, and looked for effects on Ca2+-sensing. As shown in Fig. 3
, three general phenotypes were observed. Most mutants (e.g., T889A, Fig. 3 B) showed little or no change in Ca2+ response. This was the case for 11 of the 20 mutants tested. Nine mutants showed reduced Ca2+-sensitivity, as judged by a decrease in G-V-shift in response to micromolar [Ca2+]. An example of one such mutant, D895A, whose G-V shift in response to 10 μM [Ca2+] is reduced by approximately half, is shown in Fig. 3 C. Most interesting, however, were two mutants, D898A and D900A (Fig. 3, D and E); each exhibited essentially no response to 10 μM [Ca2+]. Thus, large deletions or substitutions are not required to eliminate Ca2+ sensing via the Ca2+-bowl–related site; a single-point mutation at D898 or D900 is sufficient.
These results are summarized in Fig. 4
A, where we have plotted for each mutant tested the change in half-maximal activation voltage (ΔV1/2) observed in response to increasing [Ca2+] from 0.003 to 10 μM. When the data are viewed in this way, a trend becomes apparent. The most effective mutations involve the very acidic region between D895 and D903 with a rough progression as one moves from the center of this region toward either end. D898A and D900A are most effective, and then D897A, Q896A, and D895A are less so on the left, and similarly D901A and P902A are less so on the right. What is striking, however, is that D899A, a substitution at the very center of this acidic region, and one flanked by D898 and D900, is without effect.
To convert these results into energetic terms (Bao et al., 2002) we used a model of the BKCa channel's voltage-dependent gating mechanism developed by Horrigan and Aldrich (1999). G-V curves from each experiment were fitted with Eq. 2, which describes the open probability of the Horrigan and Aldrich (1999) model as a function of voltage (V).
Here JO(0) and JC(0) represent the equilibrium constants for voltage-sensor movement in each subunit at 0 mV when the channel is open or closed, respectively, z represents the gating charge associated with each voltage sensor, q represents the gating charge associated with the model channel's central conformational change, and (most pertinent here) ACa represents a Ca2+-dependent factor that is logarithmically related to the free energy difference between open and closed at 0 mV as follows:
Thus, from ACa, JC(0), and JO(0) we can estimate ΔG(0)O − C as a function of [Ca2+] for each mutant and then subtract to calculate ΔΔGCa(0.003–10 μM), the change in this free-energy difference as [Ca2+] is raised from 0.003 (nominally 0) to 10 μM. In so doing we constrained the fitting by using voltage-dependent gating parameters close to those determined for mSlo by Horrigan et al. (1999)(see also Cox and Aldrich, 2000), which is equivalent to assuming that Ca2+-bowl mutations do not affect voltage-sensor movement directly. This is perhaps a questionable assumption, but one we think is reasonable, given that Ca2+ binding and voltage-sensor movement affect channel opening independently (Cox et al., 1997; Cui and Aldrich, 2000; Rothberg and Magleby, 2000; Horrigan and Aldrich, 2002), and that, in general, we have observed little effect of each Ca2+-bowl mutation on the mSlo channel's G-V relation in the absence of [Ca2+] (for examples see Fig. 3). The results of this analysis are shown in Fig. 4 B, and, as is evident, they are qualitatively similar to those observed for ΔV1/2. The side chains of D898 and D900 are critical for Ca2+ sensing, and mutations around this region, except for D899A, reduce ΔΔGCa by a third to two thirds.
Aspartates Are Needed at 898 and 900 for Ca2+ Sensing
Thus, effective mutations in the Ca2+ bowl lie in a cluster that might reasonably form a Ca2+-binding site, and if this is the case, then carboxylate oxygens from residues D898 and D900 are likely to be required for Ca2+ coordination. To test this hypothesis we mutated D898 and D900 one at time to asparagines, thereby substituting in each case a neutral amide for a similarly sized carboxylate. Despite the minimal nature of these mutations, however, as is consistent with our hypothesis, they also eliminated Ca2+-bowl function (Fig. 5
C and Fig. 6
C). In addition, we obtained similar results when each of these residues was mutated individually to a positively charged lysine (Figs. 5 D and 6 D). Thus, the negatively-charged carboxylate oxygen's of D898 and D900 are essential for Ca2+-bowl function. When we mutated D898 or D900 to glutamate, however, maintaining the negative charge, Ca2+ sensitivity was also largely lost (Figs. 5 E and 6 E). The mutant channels were in each case slightly more responsive to Ca2+ than the corresponding aspartate-to-alanine mutant, but not to a statistically significant degree and much less so than was the control M513I channel. Thus, at either position glutamate's extra methylene group (as compared with aspartate) is sufficient to disturb Ca2+-bowl function. Charge, therefore, is not the only determinant of function at these residues, and in general it appears that Ca2+-dependent channel opening is very sensitive to changes in structure at residues 898 and 900. Interestingly, if the Ca2+ bowl forms a Ca2+ binding site, this degree of structural sensitivity at points of Ca2+ coordination is not unexpected. In the EF hand–type Ca2+-binding site of troponin C, for example, D–E mutations at either the X or the Y coordinating position reduce the protein's dissociation constant for Ca2+ by 10-fold or more, and each mutation also eliminates Ca2+-induced muscle-fiber responses (Babu et al., 1992, 1993).
D898 and D900 Are also Critical for Ca2+ Binding
Although the experiments described above indicate that residues in the Ca2+ bowl are important for Ca2+ sensing, they do not speak directly to the issue of whether or not Ca2+ binds to the channel's intracellular domain in general, or the Ca2+ bowl in particular. One could imagine that mutations in the Ca2+ bowl affect a Ca2+-binding site that resides a considerable distance away, perhaps in the integral-membrane domain. Thus, we employed—as others have done—a 45Ca2+ gel-overlay assay (Ngai et al., 1987; Bian et al., 2001; Braun and Sy, 2001) to look more directly at Ca2+ binding to the channel's intracellular domain. A fusion protein composed of glutathione-s-transferase and a 207–amino acid portion of mSlo's intracellular domain—a region that spans from 137 amino acids upstream to 42 amino acids downstream of the Ca2+ bowl (Braun and Sy, 2001)—was expressed in, and purified from, bacteria. It was then electrophoresed on a SDS polyacrylamide gel, blotted to nitrocellulose, and exposed to 45Ca2+. As shown in Fig. 7
A, this fusion protein binds 45Ca2+, and binding increases as more protein is loaded onto the gel—this blot was exposed to 12 μM 45Ca2+. Furthermore, binding is greatly attenuated when D898 and D900 are both mutated to alanine (Fig. 7 B). In fact, densitometric measurements of the bands in Fig. 7 B (lower panel) indicate that Ca2+ binding is inhibited by 70% when comparing the signal from 10 μg of mutant to 10 μg of wild-type protein, and 86% when comparing 20 μg of each protein. In 32 other experiments very similar results were obtained; on average Ca2+ binding was reduced by these mutations by 80 ± 2% (n = 33, SEM). Thus, Ca2+ sensing via a site that is functionally linked to the Ca2+ bowl, and Ca2+ binding to a portion of the BKCa channel's intracellular domain that includes the Ca2+ bowl, are both strictly dependent on the structures of the aspartates at positions 898 and 900.
Ca2+ Binding to Other Mutant Fusion Proteins
To further explore the correlation between Ca2+ binding and Ca2+ sensing, we made a series of other mutations in our mSlo-tail fusion protein and examined each of these as well for Ca2+ binding, again by gel-overlay. On each gel we ran the wild-type fusion protein and the D898/D900 mutant as positive and negative controls, respectively. Results of these experiments are summarized in Fig. 8
A, where each bar indicates band density as a percentage of wild-type (bar 1, far left). 10 mutations were examined. Results from a number of them correlated well with our electrophysiological data. Mutations E884A and E905A, for example, each of which showed no significant effect on Ca2+ sensing (see Fig. 4), also showed no clear effect on Ca2+ binding (Fig. 8 A compare bars 2 and 11 to bar 1). Similarly, D903A (bar 10) showed only a small effect on Ca2+ binding and no effect electrophysiologically. Furthermore, the double mutant D898E/D900E (bar 8), although it conserves the negative charges of these residues, showed substantial effect on binding, reducing band density by an average of 57 ± 7.5% (n = 4, SEM), and, as discussed above (see Figs. 5 and 6), aspartate-to-glutamate mutations at these positions also showed large effects on Ca2+ sensing.
Some of our gel-overlay results, however, were not as consistent with our electrophysiological data. While in our electrophysiological experiments the point mutations D898A and D900A each eliminated Ca2+ sensing, these mutations only partially eliminated Ca2+ binding (Fig. 8 A, bars 4 and 6; see also Fig. 9)
, and they were less effective than the double mutation D898A/D900A. Furthermore, D899A, which had no effect on Ca2+ sensing, inhibited Ca2+ binding by an amount similar to the mutations that eliminated Ca2+ sensing, D898A and D900A (Fig. 8 A, bar 5, and Fig. 9, lane 5). Also, the double mutation D899A/D901A was almost as effective at inhibiting Ca2+ binding as was D898A/D900A (Fig. 8 A, compare bars 9 and 7), even though the latter was composed of mutations that were each more effective in our electrophysiological experiments than the mutations included in the former. Thus, overall, the correlation between mutations that disrupt Ca2+ sensing and those that disrupt Ca2+ binding is not as strong as it first appeared.
Why do the mutations D898A and D900A reduce Ca2+ binding (on a percentage of wild-type basis) less than they reduce Ca2+ sensing? The answer to this question is unclear, however, recalling that Ca2+ sensing necessarily depends on there being a difference in affinity between open and closed, we might explain these results by supposing that each single mutation is sufficient to prevent a change in affinity from occurring as the channel opens at the Ca2+-bowl–related Ca2+-binding site, while both mutations together have a larger effect on the absolute affinity of the site. For example, we have estimated previously that the Ca2+-bowl–related Ca2+ binding site has an affinity for Ca2+ of ∼3.5 μM when the channel is closed and ∼0.8 μM when the channel is open (Bao et al., 2002). We do not know whether the Ca2+-bowl–related Ca2+-binding site adopts its open or closed configuration in our fusion protein, but let us suppose it adopts its open, high-affinity configuration. Then, if D898A or D900A raise the dissociation constant of the open configuration (KO) from 0.8 to 3.5 μM without altering the affinity of the closed configuration (KC), this would eliminate Ca2+ sensing in the intact channel, but it would reduce Ca2+ binding to the fusion protein, when exposed to 10 μM [Ca2+], by only a modest 20%. Thus, that some mutations affect Ca2+ sensing more than Ca2+ binding (in terms of a percentage of maximal ΔΔGCa and 45Ca2+-band density), is not altogether unexpected.
The Ca2+ Bowl May be Distorted in the Gel Overlay Assay
It is harder, however, to account for the effect of the mutation D899A on Ca2+ binding (Fig. 8 A, bar 5). This mutation had no effect on Ca2+ sensing (see Fig. 4), but it decreased Ca2+ binding by an amount (55 ± 7.8) similar to that observed with the D898A and D900A mutations. According to allosteric theory, in order for D899A to have no effect on Ca2+'s power to shift the mSlo G-V curve, it must not change the channel's Ca2+ dissociation-constant ratio KClosed/KOpen. However, in order to have some effect in the binding assay it must change at least one of these dissociation constants (the one that corresponds to the binding site's conformation in our fusion protein). Thus, to explain the effects of D899A in both assays we seemingly must suppose that this mutation reduces both KClosed and KOpen by a common factor such that the ratio KClosed/KOpen remains unchanged, but Ca2+ binding to the fusion protein, when exposed to 10 μM 45Ca2+, is reduced, because of the now lower affinity of the mutated site.
While the above explanation is not unreasonable, it relies on the idea that D899A reduces KOpen and KClosed by a common factor, an occurrence without any obvious physical origin. A more straightforward hypothesis is that Ca2+ binds to the fusion protein's Ca2+ bowl, but the Ca2+ bowl is not in a form that it normally takes in the intact channel. The Ca2+ bowl may be in some partially denatured or distorted form such that the determinants of binding are not precisely the same as they are in the intact channel. In fact, viewing the data in Fig. 8 A in this light, it appears that, if we restrict our attention to the region between residues Q896 and D901 inclusive, 45Ca2+ binding correlates fairly well simply with the charge of this region. This is illustrated by the filled circles in Fig. 8 B, where 45Ca2+-band density (as a percentage of wild-type) is plotted versus the change in net charge each mutation brings about. Open circles indicate Ca2+-bowl mutations made outside this region, where a change in charge has less effect on binding.
An interesting aspect of the gating behavior of the BKCa channel is the apparent novelty of its Ca2+-sensing mechanism. The channel's pore-forming Slo subunit contains no EF hand, no clear C2 domain, and no other motif that can be described as canonical for Ca2+ binding. Yet the channel opens in response to as little as a few hundred nanomolar Ca2+, depending on the membrane potential, so at least moderately high-affinity Ca2+-binding sites must be involved. Indeed, for several years now an acidic region termed the “Ca2+ bowl”, in the latter part of the channel's intracellular domain, has been implicated in Ca2+ sensing (Schreiber and Salkoff, 1997; Schreiber et al., 1999; Bian et al., 2001), and more recently a picture has emerged that suggests that the BKCa channel contains two types of high-affinity Ca2+-binding sites, both of which reside in the channel's intracellular domain, and one of which is functionally linked to the Ca2+ bowl (Schreiber and Salkoff, 1997; Bian et al., 2001; Bao et al., 2002; Xia et al., 2002).
Some Acidic Side Chains Are More Important than Others
Here we have presented a systematic analysis of residues in the Ca2+ bowl that are important for Ca2+ sensing. And if the Ca2+ bowl forms a Ca2+-binding site, then what we have found might be considered not unexpected. The acidic region in the center of the Ca2+ bowl (residues 895–903) is most important for Ca2+ sensing, and outside of this region, with the exception of Q910, our mutations had little effect. Thus, the side-chain of E884, although acidic, is not required for normal Ca2+ sensing, and neither are those of the eight other oxygen-containing side chains outside of this subregion—again, except for Q910. (Note that the mutation at D888 did not express well, so the importance of this residue's side chain is unclear.)
Within the acidic center of the Ca2+ bowl, however, all positions do not contribute equally. Single-point mutations of any sort at either D898 or D900 eliminate Ca2+ sensing via the Ca2+-bowl–related site, whereas mutations at many other residues in this region—including three aspartates, a glutamine, a proline, and a threonine—show moderate effects. And mutations at the acidic residues D899 and D903 are without effect. Thus, one might speculate that if the Ca2+ bowl forms a Ca2+-binding site, then it is the side chains of D898 and D900 that most closely coordinate Ca2+ and that many neighboring residues play a lesser role; that is, they may be involved in attracting Ca2+ to the site, maintaining the general structure of the region, and/or also providing coordinating ligands. In fact, the observation that the mutations D898A and D900A each eliminate Ca2+ sensing, while D899A has no effect, suggests that the Ca2+ bowl may form a binding loop with D899 at the center of the turn, its side chain extending away from the Ca2+ ion, and D898 and D900, on either side, extending their side chains inward to coordinate Ca2+.
Although clearly speculative, to test the feasibility of this hypothesis we positioned the residues of the acidic region of the Ca2+ bowl (residues 896–907) according to the backbone structure of the Ca2+ binding loop of parvalbumin (residues 51–62), and then energy minimized this configuration. Interestingly, as shown in Fig. 10
, the result of this exercise indicates that such an arrangement is possible, and further, it suggests that the backbone carbonyl oxygen of P902 may also play a role in Ca2+ coordination.
Ca2+ Binds to the Intracellular Domain
Of course Ca2+ sensing and Ca2+ binding, although intertwined, are not equivalent. Because Ca2+-dependent channel opening depends on a difference in affinity between open and closed, mutations that eliminate opening in response to Ca2+ binding need not destroy binding altogether. And if a mutation eliminates a Ca2+-binding site, it need not sit at the point of Ca2+ coordination. Thus, our electrophysiological data are suggestive, but they do not force the conclusion that the Ca2+ bowl forms a Ca2+-binding site. Indeed, given recent results that suggest that the BKCa channel retains its Ca2+ sensitivity even after its entire intracellular domain is eliminated (Piskorowski and Aldrich, 2002), indirect effects of mutations at the Ca2+ bowl seem an important consideration.
To more directly test for Ca2+ binding to the Ca2+ bowl we examined, by 45Ca2+ gel-overlay, Ca2+ binding to a series of wild-type and mutant fusion proteins consisting of GST and a portion of the Slo's COOH-terminal “tail” that includes the Ca2+ bowl. Most interesting, we found that these fusion proteins do bind Ca2+ in this assay, and, just as with Ca2+ sensing, it is in the acidic central region of the Ca2+ bowl that mutations have greatest influence on Ca2+ binding. Furthermore, when we mutated the residues that most effectively disrupted Ca2+ sensing, D898 and D900, together to alanine, we saw an 80% decrease in Ca2+ binding. Thus, there is a Ca2+ binding site within the intracellular region contained in the fusion protein we have used in our binding studies, and it seems likely that this binding site is formed by the Ca2+ bowl and relevant to Ca2+-dependent channel opening.
The Ca2+ Bowl May be Distorted in Gel-overlay
When we looked more extensively, however, at the relationship between mutations that affect Ca2+ sensing and those that affect Ca2+ binding, inconsistencies arose that suggest that the structure of the Ca2+ bowl in our gel-overlay assay may not be the same as it is in either the open or closed conformation of the intact channel. Most difficult to explain, we found that the mutation D899A had no effect on Ca2+ sensing, but it reduced Ca2+ binding by an amount similar to D898A and D900A, both of which eliminated Ca2+ sensing. And similarly, the double mutant D899A/D901A was almost as effective at reducing Ca2+ binding as was the double mutant D898A/D900A, even though the former was composed of two mutations that showed either no or only a partial effect on Ca2+ sensing, while the latter was composed of mutations that each eliminated Ca2+ sensing. In fact, taken together, our gel-overlay results suggest that in the middle of the Ca2+ bowl (895–901), the effects of mutations on Ca2+ sensing correlate better with the loss of negatively charged side chains than with the mutation of particular sidechains. Thus, it may be that in our gel-overlay assay the Ca2+ bowl takes a conformation that is different from any of its native conformations, one perhaps in which the central Ca2+-bowl aspartates contribute more equally to Ca2+ coordination. Considering that during the gel-overlay assay the protein is denatured in the gel and then renatured on the blot, this is perhaps not a surprising conclusion—although this assay has been used successfully with a great many Ca2+-binding proteins, including those with EF hand, C2, and novel types of Ca2+-binding sites (Maruyama et al., 1984; Sienaert et al., 1997; Menguy et al., 1998; Hammarberg et al., 2000; Jegerschold et al., 2000; Rajini et al., 2001; Tompa et al., 2001; Bandyopadhyay et al., 2002; Morohashi et al., 2002). Still, clearly, a different Ca2+-binding assay would be desirable. At present, however, we and others have been limited in this regard by the poor solubility of Slo-tail GST-fusion proteins.
Is the Ca2+ Bowl a Binding Site In Vivo?
The observation described above raises the issue of whether the Ca2+ bowl is a Ca2+-binding site at all in the native channel. Or does it bind Ca2+ only after being subjected to SDS-PAGE and electroblotting? From our data we cannot answer this question definitively; however, we think that our observations that the same subregion of the Ca2+ bowl is important for both Ca2+ sensing and Ca2+ binding and that Ca2+ sensing depends on two negatively charged residues separated by one less critical residue—a common component of both EF-hand and C2 type Ca2+ binding sites (Falke et al., 1994; Nalefski and Falke, 1996)—tips the balance in favor of the hypothesis that this region forms a functionally relevant Ca2+-binding site in vivo.
But if this is the case, how can we explain the results of Piskorowski and Aldrich (2002), who found that this part of the channel—and in fact the channel's entire intracellular domain—is not needed for Ca2+ sensing? The answer to this question is not apparent, although it could be that there are Ca2+-binding sites in the channel's integral-membrane domain that are unrelated to the Ca2+ bowl and responsible for the Ca2+ sensitivity observed for the severely truncated channel (Piskorowski and Aldrich, 2002). If this is the case, however, then such sites must be either uncovered upon truncation or disabled by mutations in the channel's intracellular domain, as mutations in this domain alone are sufficient to prevent Ca2+-dependent activation (Fig. 1; Bao et al., 2002; Xia et al., 2002).
Bearing on this issue, Qian and Magleby (2003) have recently reported that when a channel carrying mutations in both its RCK domain and its Ca2+ bowl, and thus insensitive to micromolar Ca2+, is expressed together with the BKCa β1 subunit, some Ca2+ sensitivity is restored to the channel. Since the β1 subunit does not appear to have Ca2+-binding sites of its own (Qian and Magleby, 2003), this result argues either that the β1 subunit restores activity at the Ca2+-binding sites disabled by the mutations, or that the β1 subunit uncovers dormant Ca2+-binding sites. Perhaps, then, Ca2+ sensing in the BKCa channel can involve dormant sites that reside in the channel's integral membrane domain. To explain our results, however, no such a hypothesis is required.
Previous Biochemical Studies
Others have found before us that Ca2+ will bind to a fusion protein that contains amino acids that span the Ca2+ bowl. Braun and Sy (2001) observed, by gel-overlay assay, 45Ca2+ binding to a recombinant mSlo peptide similar to the one we have used in our study, but they did not show that Ca2+-bowl mutations could affect this binding. Bian et al. (2001), however, did make this connection. They found that Ca2+ would bind to a fusion protein that contained a 280–amino acid portion of Drosophila Slo (also by gel-overlay) and that this binding was inhibited by 56% when the Ca2+-bowl residues 897–901 were mutated to asparagine. Thus, our results are consistent with these previous reports, although they may be distinguished in three respects. First, our most effective mutation in the binding assay D898A/D900A inhibited Ca2+ binding to a larger extent than did the 897–901N mutation of Bian et al. (2001) (80% compared with 56%). Second, Bian et al. (2001) examined only one mutation in their binding and electrophysiological assays, thus they were not likely to see the inconsistencies we have observed. And third, our mutations D898A and D900A much more effectively altered Ca2+ sensing than did theirs. In fact, in their study it was not apparent that their 897–901N mutation decreased the change in V1/2 normally observed as [Ca2+] is increased. This latter difference, however, likely arises because their mutant still contained functioning high-affinity Ca2+-binding sites of the second (RCK-associated) type (Bao et al., 2002; Xia et al., 2002), and because the Ca2+ concentrations they used were high enough to activate the channel's low-affinity Ca2+-binding sites (Shi and Cui, 2001; Zhang et al., 2001; Bao et al., 2002; Shi et al., 2002). Thus, a direct comparison is perhaps not warranted. Nevertheless, Bian et al. (2001) interpreted their data to indicate a loss of function at a Ca2+-bowl–related Ca2+-binding site, an interpretation that we think is correct.
Previous Electrophysiological Studies
As mentioned above, because we have taken steps to minimize interference from the BKCa channel's RCK-associated high-affinity Ca2+-binding sites and its low-affinity sites, it is often difficult to directly compare the electrophysiological data presented here to previous results; however, in the main, they appear consistent. Schreiber and Salkoff (1997), in the paper that defined the Ca2+ bowl, described five mutations (D898N, 897–899N, 897–901N, Δ897–898, Δ897–899) that altered channel gating in manner that they interpreted to indicate loss of function at the Ca2+ bowl. This interpretation has been further supported for the 897–901N mutant by three other studies (Bian et al., 2001; Niu and Magleby, 2002; Xia et al., 2002). All of these mutants contain mutations at 898 or at both 898 and 900, as is consistent with our data. We found in a previous study that the mutations Δ896–903, Δ898–903, and Δ899–903 could each eliminate Ca2+ bowl function (Bao et al., 2002), and again each contains a mutation at position 898 or 900 or both. And Braun and Sy (2001) found that a mutation at position 895 reduces the channel's response to Ca2+ but does not eliminate it, and we have observed a similar partial loss of Ca2+ sensitivity. No other Ca2+-bowl mutations have been reported.
Here we have presented the first systematic analysis of the side chains of the Ca2+ bowl that are most important for Ca2+ sensing, We have narrowed the functionally relevant region of the Ca2+ bowl from 28 to ∼10 central residues, and we have shown that the many acidic residues in this central region are not functionally equivalent. Also, we have found that Ca2+ will bind to a peptide composed of GST and a 207–amino acid part of the mSlo tail that includes the Ca2+ bowl and that mutations that eliminate Ca2+ sensing via the Ca2+-bowl–related site also greatly attenuate Ca2+ binding. Inconsistencies between our binding and electrophysiological data, however, have caused us to question whether the Ca2+ bowl is in a native conformation in our gel-overlay assay, and thus they suggest caution in using this assay to study Ca2+ binding at the Ca2+ bowl.
We gratefully acknowledge Dr. Andrew Braun for the generous gift of his GST-mSlo fusion-protein construct as well as a great deal of technical advice, Dr. Jim Baleja for his structural modeling of the Ca2+ bowl (Fig. 10), Dr. Christopher Schmid for statistical advice, Dr. Christopher Miller for helpful discussions, and Dr. Kathleen Dunlap, Dr. Robert Blaustein, and Anne Rapin for helpful comments on the manuscript.
This work was supported by grant R01HL64831 from the National Institutes of Health and by a grant from The Jessie B. Cox Charitable Trust and The Medical Foundation.
Olaf S. Andersen served as editor.