Structure and function of multiple Ca²⁺-binding sites in a K⁺ channel regulator of K⁺ conductance (RCK) domain

Abstract

Regulator of K⁺ conductance (RCK) domains control the activity of a variety of K⁺ transporters and channels, including the human large conductance Ca²⁺-activated K⁺ channel that is important for blood pressure regulation and control of neuronal firing, and MthK, a prokaryotic Ca²⁺-gated K⁺ channel that has yielded structural insight toward mechanisms of RCK domain-controlled channel gating. In MthK, a gating ring of eight RCK domains regulates channel activation by Ca²⁺. Here, using electrophysiology and X-ray crystallography, we show that each RCK domain contributes to three different regulatory Ca²⁺-binding sites, two of which are located at the interfaces between adjacent RCK domains. The additional Ca²⁺-binding sites, resulting in a stoichiometry of 24 Ca²⁺ ions per channel, is consistent with the steep relation between [Ca²⁺] and MthK channel activity. Comparison of Ca²⁺-bound and unliganded RCK domains suggests a physical mechanism for Ca²⁺-dependent conformational changes that underlie gating in this class of channels.

Regulator of K⁺ conductance (RCK) domains are structurally conserved ligand-binding domains that control the activity of a diverse array of K⁺ channels and transporters (1–3). Many prokaryotic RCK domains contain a conserved sequence motif for binding of nucleotides (NAD⁺ or ATP) (4, 5). In some prokaryotic and most of the known eukaryotic RCK-containing K⁺ channels, however, the nucleotide binding motif is absent, and these channels are modulated by cytoplasmic ions such as Na⁺, H⁺, or Ca²⁺ (6–12).

MthK is a prototypical RCK-containing K⁺ channel that has provided insight toward the structural basis of ion channel gating by RCK domains (2, 13, 14). In MthK, binding of Ca²⁺ to an octameric ring of RCK domains (the gating ring), which is tethered to the pore of the channel, leads to a series of conformational changes that facilitates channel opening and K⁺ conduction (2, 15, 16). Based on X-ray structures of the Ca²⁺-bound MthK channel and the unliganded MthK gating ring (17), it has been hypothesized that the principal Ca²⁺-dependent conformational change is initiated by the movement of a Glu side chain (E212) at a single Ca²⁺-binding site within each RCK domain (Fig. 1A; site 1, formed by D184, E210, and E212), followed by subsequent movement of a nearby Phe side chain (F232). However, the conformational changes in the immediate vicinity of site 1 are relatively subtle compared to apparent conformational changes in other regions of the RCK domains (17); thus the mechanism by which Ca²⁺ binding at site 1 modulates channel gating is unclear.

Fig. 1.

Charge-neutralization of residues at site 1 does not eliminate Ca²⁺ activation of MthK. (A) Ca²⁺ coordination at site 1. (B) Representative currents through MthK WT and mutant channels with 4 mM Ca²⁺, pH 7.7 at the cytoplasmic side of the membrane. Each of these bilayers contained three active channels, except for the D184A trace, which contained one active channel. D184N, D184A, E210Q, and the D184N/E210Q double mutant resulted in decreased opening at this [Ca²⁺], compared to WT and E212Q. (C) Po vs. [Ca²⁺] relations for MthK WT and mutants. Each of these mutants could be activated to the same maximal Po as WT; thus neutralization of D184 and E210 could decrease affinity for Ca²⁺, but neither mutation alone or combined could eliminate Ca²⁺ activation, consistent with the presence of an additional Ca²⁺ binding site. Data points represent mean ± SEM of bilayer experiments in which data were obtained at six different [Ca²⁺]. WT, n = 5; D184N, n = 5; D184A, n = 5; E210Q, n = 6; E212Q, n = 8; D184N/E210Q, n = 6.

To gain insight toward mechanisms underlying Ca²⁺-dependent conformational changes in the MthK RCK domain, we probed MthK structure and function using electrophysiology and crystallography. Our results demonstrate that whereas site 1 contributes energetically to Ca²⁺-dependent gating, charge-neutralization of the key Ca²⁺-coordinating residues at this site do not eliminate Ca²⁺-dependent gating. Examination of crystallographic data revealed two additional potential Ca²⁺-binding sites at RCK subunit interfaces. Charge-neutralizing mutations of Ca²⁺-coordinating side chains at each of these sites disrupt Ca²⁺-dependent gating, and combined mutation at all three sites effectively eliminates Ca²⁺-dependent activation of the channel at [Ca²⁺] up to 100 mM. Comparison of the unliganded and Ca²⁺-bound structures at the Ca²⁺-binding sites further reveals the elementary steps in intersubunit Ca²⁺-bridge formation, elucidating a physical mechanism by which Ca²⁺ might stabilize RCK domain conformation to facilitate channel opening.

Results

Charge-Neutralizing Mutations at Site 1 do not Eliminate Ca²⁺-Dependent Activation of MthK.

MthK opening is facilitated by Ca²⁺ binding at site 1, which is formed by the side chains of D184, E210, and E212 (Fig. 1A) (2, 18). To determine the energetic roles of Ca²⁺-coordinating residues at site 1 in MthK gating, we assessed the effects of charge-neutralizing mutations at site 1 using single-channel recording of purified MthK channels incorporated into planar lipid bilayers. Although the mutations D184N and E210Q (but not E212Q) significantly increased the EC₅₀ for Ca²⁺ activation (i.e., decreased the Ca²⁺ sensitivity) (Fig. 1B, Table S1), these mutant channels could still be activated by Ca²⁺ to maximal open probability (Po) > 0.9, with steep Hill coefficients that were not significantly different from WT (Fig. 1C, Table S1). In addition, the double mutant D184N/E210Q, which effectively eliminates the critical negative charges at site 1, does not further weaken Ca²⁺ activation compared to the E210Q or D184N single mutants (Fig. 1C, Table S1). In addition, the mutation D184A, which eliminates the oxygen-containing side chain at this position, weakens Ca²⁺-dependent activation to an extent similar to the D184N and E210Q mutations (Fig. 1C, Table S1). These results are consistent with the importance of the D184 and E210 carboxylates in Ca²⁺ binding and MthK channel activation. However, the persistence of Ca²⁺-dependent activation despite elimination of the negative charges seemed at odds with site 1 being the sole determinant of Ca²⁺ activation in MthK, and suggested that there might be an additional site (or sites) contributing to Ca²⁺-dependent gating.

MthK RCK Domains Contain Two Additional Ca²⁺-Binding Sites Located at the Flexible Interface.

To determine whether MthK might contain additional Ca²⁺-binding sites that had not been identified previously, we initially probed the electron density map generated using the published structure factors deposited with Protein Data Bank (PDB) ID code 1LNQ, the structure of the MthK channel crystallized in the presence of Ca²⁺ (2). We examined these data after reasoning that electron density peaks corresponding to single cations within the large MthK RCK domain may have been difficult to unambiguously identify with low-resolution crystallographic data, and thus may not have been modeled in the 1LNQ structure.

From electron density maps generated with the 1LNQ structure factors, we identified two candidate peaks (per RCK domain) of relatively strong electron density (Fig. S1). Both peaks were located at the confluence of at least two acidic side chains in the MthK RCK domain, and had relative amplitudes of > 5σ in the 2F_o - F_c map, and amplitudes of > 3.5σ in the F_o - F_c map. These identified peaks were located at the flexible interface between pairs of RCK domains, with one peak between the side chain carboxylates of D245 and E248 of the same RCK subunit, and the other between the side chain carboxylates of D305 in one subunit and E326 of the adjacent subunit. In the 1LNQ structure, the D245/E248 and D305/E326 side chains each appear to be interacting with one another via hydrogen bonds. Hydrogen bonding between these side chains is a plausible idea, if we assume that these acidic side chains are protonated at pH 6.5 (the approximate pH during growth of this crystal), and thus would not electrostatically repel one another. However, an alternative hypothesis is that these side chains are deprotonated and thus negatively charged, and the observed density could be accounted for by a cation that would provide a stabilizing countercharge. As illustrated in Fig. S1, the peaks > 5σ in the 2F_o - F_c map were stronger than the other modeled density in these regions, and thus appeared to be good candidates for the location of electron-dense Ca²⁺ ions.

To further confirm the locations of these electron density peaks, we grew additional crystals of the MthK channel in the presence of Ca²⁺ and analyzed diffraction datasets from the best diffracting crystals. We then probed in detail the structure of one crystal that diffracted to 3.4 Å, building Ca²⁺ ions at the positions of the strong electron density peaks identified both in this best-diffracting dataset and in the 1LNQ dataset (PDB ID code 3RBZ) (Table 1).

Table 1.

Data collection and refinement statistics

	MthK, full-length, +Ca2+	MthK RCK D184N, +Ca2+
Data collection
Space group	P6122	P6522
Cell dimensions
a, b, c (Å)	138.6, 138.6, 371.6	119.1, 119.1, 351.1
α, β, γ (°)	90, 90, 120	90, 90, 120
Resolution (Å)	48.5–3.4 (3.52–3.39)	49.5–2.8 (2.95–2.80)
Rsym	0.132 (0.421)	0.104 (0.468)
I/σI	15.0 (2.1)	13.6 (3.7)
Completeness (%)	93.6 (56.6)	99.4 (96.5)
Redundancy	11.3 (5.8)	11.3 (7.4)
Refinement
Resolution (Å)	48.5–3.4	49.5–2.8
No. reflections	26,324	36,614
Rwork/Rfree	25.7/28.8	24.0/27.4
No. atoms
Protein	8,142	10,043
Ligand/ion	12	12
Water	0	0
B-factors
Protein	96.9	55.2
Ligand/ion	60.2	66.5
r.m.s. deviations
Bond lengths (Å)	0.012	0.009
Bond angles (°)	1.27	1.05

Values in parentheses are for highest-resolution shell.

The locations of the putative Ca²⁺-binding sites in the MthK RCK domain are shown in Fig. 2. The first site, site 2, contains a Ca²⁺ ion coordinated with an apparent bipyramidal geometry, consistent with that of many known Ca²⁺-binding sites, including site 1 (Fig. 1A) (18, 19). After building a Ca²⁺ ion at site 2 and further refining the structure, the Ca²⁺ ion is coordinated by two oxygens from the E248 side chain (mean Ca²⁺-O distances, 2.5 and 2.7 Å) and the main chain carbonyl oxygen of D245 (mean Ca²⁺-O distance, 2.3 Å), which form a plane with the coordinated Ca²⁺. The capstones of an apparent bipyramid at site 2 are formed by protein oxygens from the E266 side chain (mean Ca²⁺-O distance, 2.4 Å) and the main chain carbonyl of R241 (mean Ca²⁺-O distance, 2.9 Å) (Fig. 2B).

Fig. 2.

Locations of additional Ca²⁺-binding sites in the MthK RCK gating ring. (A) Gating ring of the crystallized MthK channel with additional Ca²⁺ ions built in the structure, viewed from above; the transmembrane domains have been omitted for clarity. Putative binding sites were identified as described in Results; positions of Ca²⁺ ions are indicated by green spheres. (B) Ca²⁺ coordination at site 2. Ball-and-stick representation of Ca²⁺ and protein at site 2 , with dashed lines to indicate coordination by nearby protein oxygens (Left); simulated-annealing omit map, contoured at 1.5σ, overlayed on the model (Middle); and F_o - F_c difference map calculated with Ca²⁺ omitted from the model, contoured at 10σ, overlayed in the same region (Right). (C) Ca²⁺ coordination at site 3 (Left); simulated-annealing omit map contoured at 1σ, overlayed on the model (Middle); and F_o - F_c map contoured at 10σ, overlayed in the same region (Right).

The second binding site, site 3, was formed by the side chain of D305 (mean Ca²⁺-O distances, 2.6 and 2.8 Å) and the main chain carbonyl oxygen of G290 (mean Ca²⁺-O distance, 2.9 Å), and the side chain of E326 from the adjacent subunit (mean Ca²⁺-O distances, 2.5 and 2.9 Å). At site 3, each of the side chain oxygens from D305 and E326 lie approximately in a plane with the Ca²⁺ ligand, and the carbonyl oxygen of G290 lies normal to the plane, thus forming a single pyramid (Fig. 2C, Left). Because of the relatively high B-factors for this low-resolution structure, interpretation of Ca²⁺-O coordination distances provided should be approached with caution; however, the distances that estimated from the atomic coordinates in this model conform to those typically observed for Ca²⁺ coordination in proteins and small molecules (19, 20).

Crystal Structure of the D184N Mutant RCK Domain in the Presence of Ca²⁺ is Consistent with Ca²⁺-Activation Pathways Outside of Site 1.

We have observed that the D184N mutation, at site 1, does not completely abolish Ca²⁺-dependent gating, and combining this mutation with the additional site 1 mutation E210Q does not further decrease Ca²⁺-dependent gating (Fig. 1). Although these observations are consistent with the presence of additional Ca²⁺-binding sites outside of site 1, it is alternatively possible that neither the D184N mutation nor the D184N/E210Q double mutation completely abolish Ca²⁺ binding. If this possibility were the case, then Ca²⁺ should bind at a D184N-mutated site 1, and this binding should be detectable in crystallographic data if the [Ca²⁺] were sufficiently high (i.e., high enough to activate the channel in electrophysiological experiments). We tested this possibility by crystallizing the MthK RCK domain with the D184N mutation in the presence of 100 mM Ca²⁺; in electrophysiological terms, this [Ca²⁺] is approximately 15 times the EC₅₀ for the D184N mutant channel at pH 7.7 (Fig. 1).

The D184N RCK domain crystallized in space group P6₅22, with three dimeric RCK domains per asymmetric unit, and the structure was solved to 2.8 Å resolution. We found that in this structure, Ca²⁺ was not bound at the mutated site 1 (Fig. 3B). However, electron density peaks (modeled as Ca²⁺ ions) were observed at the putative sites 2 and 3, as observed in structures of the Ca²⁺-bound MthK channel (Fig. 3 C–D). We confirmed that D184N channels could be activated with [Ca²⁺] and pH similar to those used for crystallization (Fig. S2). Thus it appears that the Ca²⁺ ions responsible for activation of the D184N mutant act through sites 2 and 3.

Fig. 3.

The D184N mutation prevents Ca²⁺ binding at site 1, but not at sites 2 and 3. Structures of sites 1, 2, and 3 in the D184N RCK domain crystallized in the presence of 100 mM Ca²⁺. (A) Ribbon representation of a Ca²⁺-bound D184N RCK dimer. Key side chain and main chain atoms are shown in ball-and-stick representation, and Ca²⁺ ions are indicated by green spheres. (B) Structure of the D184N mutant site 1 (Left), plus overlay with 2F_o - F_c map contoured at 1.2σ (Right, blue mesh). Density corresponding to a Ca²⁺ ion is not observed at the mutant site 1. (C) Site 2 overlayed with simulated-annealing Ca²⁺ omit map contoured at 1σ (Left, blue mesh), and F_o - F_c difference map calculated with Ca²⁺ omitted from the model, contoured at 6σ (Right, green mesh). (D) Site 3 overlayed with simulated-annealing Ca²⁺ omit map contoured at 1σ (Left, blue mesh), and F_o - F_c difference map calculated with Ca²⁺ omitted from the model, contoured at 6σ (Right, green mesh). These results suggest that Ca²⁺ can bind at sites 2 and 3 independent of detectable Ca²⁺ binding to site 1.

Although the D184N RCK domain did not crystallize in the biological gating ring form under these conditions (Fig. S3), Ca²⁺ coordination at sites 2 and 3 is nearly identical to that observed in the Ca²⁺-bound channel structure, which is in the gating ring form (Fig. 2 and Fig. S4). In addition, crystallization in the dimeric form would not be expected to preclude the binding of Ca²⁺ at site 1, as Ca²⁺ bound at site 1 is observed in the dimeric high-resolution MthK RCK domain structure (PDB ID code 2AEF) (18).

The structure of the Ca²⁺-bound D184N RCK domain suggests that Ca²⁺ can bind to sites 2 and 3 in the MthK RCK domain, independent of Ca²⁺ binding to site 1; thus this crystallographic analysis is further consistent with the idea that Ca²⁺ binding to sites 2 and 3 may serve as alternative pathways to activate the MthK channel.

Sites 2 and 3 Contribute Functionally to Ca²⁺-Dependent Gating.

Although sites 2 and 3 are consistent with Ca²⁺-binding sites in crystallographic data from the MthK channel and D184N RCK domain in the presence of Ca²⁺ (Fig. 2 and Fig. 3), it is important to determine whether these sites behave as Ca²⁺-binding sites that facilitate MthK channel activation. To test whether sites 2 or 3 contribute functionally to Ca²⁺-dependent gating of MthK, we recorded current through MthK channels with charge-neutralizing mutations at each of the side chains at these sites that appear to coordinate Ca²⁺ directly (E248Q and E266Q in site 2, and D305N and E326Q in site 3) (Fig. 4). As with mutations at site 1 (Fig. 1), each of these mutations significantly increased the EC₅₀ for Ca²⁺ activation (mean EC₅₀’s± standard deviation were as follows: WT, 2.2 ± 0.3 mM; E248Q, 5.0 ± 0.1 mM; E266Q, 3.2 ± 0.6 mM; D305N, 3.4 ± 0.4 mM; E326Q, 18.5 ± 0.7 mM), suggesting that each of the acidic residues in sites 2 and 3 contribute to Ca²⁺-dependent gating in MthK (Fig. 4B). In terms of the size of the shift in EC₅₀ resulting from each mutation, these followed the series E326Q > D184N > E248Q > E210Q ≈ D305N ≈ E266Q.

Fig. 4.

Mutations at sites 2 and 3 reduce Ca²⁺ sensitivity of MthK channels. (A) Representative currents through MthK mutant channels. Each of the single mutant traces contain one active channel; the D184N/E326Q trace contains four active channels, and the D184N/E248Q/E326Q mutant contains an unknown number of channels. The D184N/E248Q/E326Q trace shows the only burst observed in 5 min of continuous recording for this bilayer. Vertical scale bar is 20 pA for D184N/E326Q trace, 10 pA for all others. (B) Po vs. [Ca²⁺] for MthK WT and mutants at sites 1, 2, and 3. D184N (site 1), E248Q and E266Q (site 2), and D305N and E326Q (site 3) channels showed decreased Ca²⁺ sensitivity compared to WT. (C) Po vs. [Ca²⁺] for MthK WT, D184N, and double mutant channels. Each double mutant further decreased Ca²⁺ sensitivity of the channel compared to their single mutant counterparts. Curves in B and C represent fits with a Hill equation; means of fitted parameters (determined from fitting Po vs. [Ca²⁺] data from individual bilayers) are listed in Table S1. Data points in B and C represent mean ± SEM. WT, n = 5; D184N, n = 5; E248Q, n = 4; E266Q, n = 5; D305N, n = 4; E326Q, n = 4; D184N/E248Q, n = 3; D184N/E266Q, n = 3; D184N/D305N, n = 3; D184N/E326Q, n = 3.

Although charge-neutralization can lead to decreases in binding affinity at these putative binding sites, an effect on Ca²⁺-binding affinity is likely to be only one of several possible mechanisms underlying the functional consequences of each mutation. Because Ca²⁺ binding may be associated with movement of Ca²⁺-coordinating side chains, one or more of these mutations may disrupt both Ca²⁺ binding and the allosteric coupling between Ca²⁺ binding and channel opening, at a single binding site. In addition, a single mutation may affect Ca²⁺ activation through multiple sites, because Ca²⁺ activation through different sites may be energetically coupled to one another.

To begin to explore whether mutations at the putative sites 2 or 3 might act solely through disrupting the coupling between Ca²⁺ binding at the established site 1 and gating of the channel, we first reasoned that if a site 2 or 3 mutation were acting only through disrupting coupling of site 1 to channel gating, then combining the site 2 or 3 mutation with the D184N mutation (which effectively knocks out Ca²⁺ binding to site 1; Fig. 3B) should not further decrease Ca²⁺ sensitivity of the channel compared to the effect of the D184N mutation alone. Thus we tested whether mutations at the putative sites 2 or 3 might act solely through disrupting the coupling between Ca²⁺ binding at site 1 by generating a series of double mutants in which single mutations at sites 2 or 3 were combined with the D184N mutation and analyzing their Ca²⁺ sensitivity by electrophysiology.

In each case, the second (combined) mutation augmented the effect of the D184N mutation on EC₅₀ by several fold (Fig. 4C). For example, the D184N mutation alone increased the EC₅₀ by 2.9-fold compared to WT (increased from 2.2 to 6.3 mM Ca²⁺); additionally incorporating the E248Q, E266Q, D305N, or E326Q mutations resulted in 2.9-, 3.0-, 4.6-, and 7.6-fold increases in EC₅₀ over the D184N mutant, respectively (mean EC₅₀’s± standard deviation were as follows: D184N/E248Q, 18.0 ± 3.3 mM; D184N/E266Q, 18.7 ± 2.0 mM; D184N/D305N, 28.7 ± 1.9 mM; D184N/E326Q, 47.8 ± 5.9 mM). These results, combined with our structural data, are consistent with the idea that E248, E266, D305, and E326 act through a Ca²⁺ activation mechanism that is separate from site 1, and do not act solely by modulating gating through site 1.

Because none of the double mutations tested above completely eliminated Ca²⁺-dependent activation of MthK, we wondered whether sites 1, 2, and 3 could account for all of the Ca²⁺-dependent activation. To test this idea, we generated a triple mutant, D184N/E248Q/E326Q, which neutralized key residues at each of the putative Ca²⁺ sites. Over many repeated attempts at channel incorporation (> 50) using [Ca²⁺] ranging from 5 to 100 mM, we were able to detect only very low levels of activity using the triple mutant channel. An example of this activity is shown in Fig. 4A (similar results were observed in four additional bilayers). Overexpression and purification of the D184N/E248Q/E326Q triple mutant was indistinguishable from that of other mutant and WT MthK channels, as indicated by SDS-PAGE and gel-filtration chromatography (Fig. S5); thus it appears that the combined D184N/E248Q/E326Q mutations were sufficient to eliminate Ca²⁺-dependent gating over the range of [Ca²⁺] examined, up to 100 mM. The ability of the D184N/E248Q/E326Q triple mutation to eliminate Ca²⁺-dependent gating with [Ca²⁺] < 100 mM is consistent with the idea that sites 1, 2, and 3 could account for all of the Ca²⁺-dependent activation of MthK.

Discussion

Each MthK RCK Domain Contributes to Three Ca²⁺-Binding Sites.

Our X-ray and electrophysiological experiments are consistent with the presence of three Ca²⁺-binding sites for each RCK domain of the MthK channel. One of these sites (site 1, Fig. 1A) had been identified in previous crystallographic studies (2, 18), and the decreased Ca²⁺ sensitivity observed with the D184N and E210Q mutations is consistent with a role for Ca²⁺ binding at site 1 in MthK gating (Fig. 1). Interestingly, charge-neutralization of E212 slightly increased apparent Ca²⁺ sensitivity. The E212 side chain does participate in Ca²⁺ coordination, as indicated by distances between Ca²⁺ and an oxygen from the E212 side chain carboxylate of 2.3 to 2.4 Å, and seems to initiate a key Ca²⁺-dependent conformational change in the RCK domain (2, 17, 18) (Fig. 1 and Fig. 5). Although the mechanism underlying the effect of the E212Q mutation is not yet clear, it is possible that the effect may result from some combination of effects on coupling between Ca²⁺ binding and gating, biasing of intrinsic gating toward the open state, and/or effects on Ca²⁺-binding affinity (21).

Fig. 5.

Apparent Ca²⁺-dependent movements in MthK RCK domains. (A) Structures of Ca²⁺-free (Left, cyan; PDB ID code 2FY8) and Ca²⁺-bound RCK dimers (Right, magenta; PDB ID code 3RBZ); dashed arrows indicate the relative domain motions that follow Ca²⁺ binding. (B) Alignment of structures in A by the main chain atoms of residues 120–230 (encompassing site 1). Ca²⁺ binding results in movement of the E212 side chain, which is thought to affect interactions between F232 side chains in adjacent subunits. (B) Alignment of structures in A by the main chain atoms around site 2 (residues 240–250). Ca²⁺ binding results in movement of the E248 side chain, and relative movement of the E266′ side chain, with movement of the associated main chain (′ denotes residue in the adjacent subunit). (C) Alignment of structures in A by the main chain atoms around site 3 (residues 289–325). Ca²⁺ binding results in movement of the D305′ side chain from a position of apparent hydrogen bonding with H286.

Electrophysiological studies have demonstrated a very steep (> 8th power) relation between [Ca²⁺] and MthK activity, which is inconsistent with the presence of only eight Ca²⁺-binding sites per channel (16, 21). In addition, Ca²⁺-dependent activation persists in D184N, D184A, and E210Q single mutant and D184N/E210Q double mutant MthK channels, in which the primary Ca²⁺-coordinating residues at site 1 had been charge-neutralized (Fig. 1). These observations provided a rationale for further analysis of crystallographic data of the Ca²⁺-bound MthK RCK domain. Sites 2 and 3 corresponded to strong regions of electron density that, in the context of local chemistry, could be accounted for by Ca²⁺ ions (Fig. 2 and Fig. S1). Although site 3 had not previously been thought to be a Ca²⁺-binding site, it was determined as a Cd²⁺-binding site in the Cd²⁺-bound MthK RCK domain (22). Because Cd²⁺ is a potent activator of MthK channels (22, 23), it is possible that Ca²⁺ and Cd²⁺ share a common structural activation mechanism in these channels.

The identification of sites 2 and 3 in the crystal structures of the MthK channel and RCK domain (Fig. 2 and Fig. 3) yields an overall stochiometry of 24 Ca²⁺ per MthK channel, and the presence of these Ca²⁺-binding sites provides an explanation for the very steep Ca²⁺-dependence of MthK gating (16, 21). This steep Ca²⁺-dependence of gating is also consistent with a form of positive cooperativity, either among Ca²⁺-binding sites or among RCK subunits, in MthK activation (16). In the cases of sites 2 and 3, Ca²⁺ itself forms a metal bridge between adjacent RCK subunits, although it is not yet clear whether these intersubunit interactions result in energetic coupling among binding sites.

A Working Hypothesis for Ca²⁺-Dependent Conformational Changes in the RCK Gating Ring.

The locations of sites 2 and 3 at the flexible interfaces in the MthK gating ring (Fig. 2A) reveals regions where Ca²⁺ binding may exert forces that underlie channel gating [Fig. 5; unliganded conformation is from the closed MthK gating ring (PDB ID code 2FY8)] (17). Comparison of unliganded and Ca²⁺-bound RCK domain dimers illustrates the closer proximity of adjacent RCK domains at sites 2 and 3, combined with a wider gap between adjacent RCK domains near site 1, in the Ca²⁺-bound form. Alignments of each site in the unliganded and Ca²⁺-bound forms reveal the local structural consequences of Ca²⁺ binding (Fig. 5 B–D, Fig. S6). Ca²⁺ binding at site 1 results in movement of the E212 side chain (Fig. 5B); direct local movement of the main chain at site 1 is subtle, though in previous work it was hypothesized that movement of E212 influences intersubunit interactions through a nearby Phe residue (F232) (17). In contrast, Ca²⁺ binding at sites 2 and 3 result in apparent side chain movements (E248 in Fig. 5C) and associated relative movements of the main chain at the adjacent subunit (E266′ in Fig. 5C, and D305′ in Fig. 5D). In the unliganded RCK domain structure, the carboxylate group of the D305 side chain appears to be involved in a hydrogen bond with the side chain of H286 in the adjacent subunit, and this hydrogen bond is broken to permit coordination of Ca²⁺ at site 3 (Fig. S6, Fig. S7, and SI Discussion). Thus the strength of the interaction between D305 and H286 may be an important factor to modulate Ca²⁺ binding at site 3.

Relation to Ca²⁺-Dependent Gating.

An interesting feature common among sites 1, 2, and 3 is that each site is at least partially preformed, with 2–3 oxygens at each site in the Ca²⁺-free structure (PDB ID code 2FY8) occupying positions that are conserved in the Ca²⁺-bound structure (PDB ID code 3RBZ). We hypothesize that Ca²⁺ can bind to these partial sites, with completion following movement of the remaining parts of the site, and stabilization of the Ca²⁺-bound state by electrostatic interactions between Ca²⁺ and the protein oxygens. In terms of an overall conformational change, the metal bridges formed by Ca²⁺ act to glue together each pair of C-terminal subdomains at site 3, bind these C-terminal subdomains with the helix-turn-helix (alpha-F/alpha-G) regions of the neighboring subunit at site 2, and apparently stabilize individual N-terminal subdomains at site 1. These combined interactions, in the context of the RCK gating ring, likely lead to an increase in mechanical tension on the linker between the RCK domain and pore-lining helices to stabilize a conducting state of the channel, as hypothesized previously (17). Further experiments will be required to establish the structural basis of coupling between the RCK domains and gating of the MthK pore.

Materials and Methods

Channel Purification and Reconstitution.

Protein was expressed in Escherichia coli XL-1 blue cells on induction with 0.4 mM IPTG, and purified by metal affinity chromatography as described previously (21). Channels were reconstituted into liposomes composed of E. coli lipids (Avanti) (24), and the proteoliposomes were rapidly frozen in liquid N₂ and stored at -80 °C until use. Mutations were generated using the QuickChange Site-Directed Mutagenesis Kit (Stratagene) and confirmed by DNA sequencing.

Lipid Bilayer Recording and Solution Changes.

Recordings were obtained using planar lipid bilayers of 1-palmitoyl-2-oleoyl-phosphatidylethanolamine (POPE) and 1-palmitoyl-2-oleoyl-phosphatidylglycerol (POPG), combined in a 3∶1 ratio in a horizontal bilayer chamber, at 22–24 °C. Solution in the cis (top) chamber contained 200 mM KCl and 10 mM Hepes, pH 7.0; solution in the trans (bottom) chamber contained 200 mM KCl, 10 mM Hepes, pH 7.7, and CaCl₂ at the indicated concentrations. Recordings analyzed for this work were obtained at -100 mV transmembrane voltage. Within each bilayer, multiple solution changes were performed using a gravity-fed perfusion system, and to ensure completeness of solution changes, the trans chamber was washed with a minimum of 10 chamber-volumes of solution prior to recording under a given set of conditions.

Currents were amplified using a Dagan PC-ONE patch clamp amplifier with low-pass filtering to give a final effective filtering of 1 kHz (dead time of 0.18 ms), and sampled by computer at a rate of 10 kHz. Currents were analyzed by measuring durations of channel openings and closings at each current level by 50% threshold analysis, using pClamp9.

Crystallization.

Crystallization of MthK channels was performed essentially as described previously (2). Briefly, MthK channels, with the M107I mutation that suppresses expression of soluble RCK domain subunits, were expressed and purified as described above, except that buffer used in the final gel-filtration step contained 5 mM lauryldimethylamine-oxide (LDAO, Anatrace) instead of 5 mM n-decyl-beta-d-maltoside. Purification of the MthK RCK domain containing the D184N mutation was performed essentially as described previously (17), except that we used an expression construct in which the cDNA for the transmembrane region of the MthK channel (residues 1,106) had been deleted by QuickChange mutagenesis. The crystal used in data collection formed in a 2 μL hanging drop composed of a 1∶1 mixture of protein solution and 22% PEG3350, 100 mM MES, pH 5.9, and 100 mM CaCl₂.

X-ray Data Collection and Analysis.

Diffraction data for the MthK channel were collected at beamlines 8.2.2 and 5.0.2 at the Advanced Light Source, at -180 °C under a nitrogen stream. Data from MthK channel crystals (M107I or M107I/E92A/E96T/G85A) were collected at Advanced Light Source beamline 5.0.2 at a wavelength of 1.0 Å, and were processed and scaled using HKL2000 (25). The structure was solved in space group P6₁22 by molecular replacement using PHASER (26), and refined through cycles of manual rebuilding in COOT (27, 28) and automated refinement using the PHENIX program suite (29). Iterative rebuilding and the use of fourfold non-crystallographic symmetry (NCS) restraints resulted in refinement to an R and R-free of 25.7% and 28.7%, respectively.

Diffraction data for the D184N RCK domain with Ca²⁺ were collected at beamline X-25 at the National Synchrotron Light Source, at -180 °C under a nitrogen stream, at a wavelength of 1.1 Å. These data were processed and scaled using MOSFLM and Scala, and phasing was done by molecular replacement with PHASER using the Ca²⁺-free MthK RCK domain dimer (PDB ID code 2AEJ) (18) as the search model, in space group P6₅22. Three RCK domain dimers were found in the asymmetric unit; thus model building and refinement (using COOT and PHENIX-refine) were facilitated by sixfold NCS averaging, and strong NCS constraints were maintained throughout refinement. The model was refined to an R and R-free of 23.9% and 27.4%, respectively. Figures were prepared using PyMOL.