Crystal structure of the Ca_V2 IQ domain in complex with Ca2+/calmodulin: High-resolution mechanistic implications for channel regulation by Ca²⁺

SUMMARY

Calmodulin (CaM) regulation of Ca²⁺ channels is central to Ca²⁺ signaling. Ca_V1 versus Ca_V2 classes of these channels exhibit divergent forms of regulation, potentially relating to customized CaM/IQ interactions among different channels. Here, we report the crystal structures for the Ca²⁺/CaM—IQ domains of both Ca_V2.1 and Ca_V2.3 channels. These highly similar structures emphasize that major CaM contacts with the IQ domain extend well upstream of traditional consensus residues. Surprisingly, upstream mutations strongly diminished Ca_V2.1 regulation, whereas downstream perturbations had limited effects. Furthermore, our Ca_V2 structures closely resemble published Ca²⁺/CaM—Ca_V1.2 IQ structures, arguing against Ca_V1/2 regulatory differences based solely on contrasting CaM/IQ conformations. Instead, alanine scanning of the Ca_V2.1 IQ domain, combined with structure-based molecular simulation of corresponding CaM/IQ binding energy perturbations, suggests that the C-lobe of CaM partially dislodges from the IQ element during channel regulation, allowing exposed IQ residues to trigger regulation via isoform-specific interactions with alternative channel regions.

INTRODUCTION

Voltage-gated Ca²⁺ channels of the Ca_V1-2 family constitute dominant sources of Ca2+ entry that trigger numerous biological functions (Dolmetsch, 2003; Dunlap et al., 1995). Fitting with their salient functional roles, these channels are subject to extensive positive and negative Ca²⁺ feedback control of their opening, with calmodulin (CaM) serving as the Ca²⁺ sensor (Evans and Zamponi, 2006). In particular, for Ca_V2 channels which predominate in triggering CNS neurotransmitter release (Wheeler et al., 1994), these feedback regulatory systems are believed to influence short-term synaptic plasticity (Borst and Sakmann, 1998; Chaudhuri et al., 2005; Cuttle et al., 1998; Wykes et al., 2007; Xu and Wu, 2005), and thereby the computational repertoire of the brain (Abbott and Regehr, 2004; Tsodyks and Markram, 1997). In all of these contexts, CaM clearly exhibits powerful and unexpected forms of Ca²⁺ decoding (Dunlap, 2007), the principles of which may generalize to numerous Ca²⁺ signaling systems. As such, the CaM regulation of Ca²⁺ channels not only has specific consequences for Ca²⁺ signaling, but may also represent a general mechanistic paradigm for Ca²⁺ decoding and ion-channel modulation.

CaM/Ca²⁺ channel regulation includes several provocative functional features. First, in resting channels, Ca²⁺-free CaM (apoCaM) is already preassociated with the cytoplasmic region of channels, rendering CaM as a ‘resident’ Ca²⁺ sensor (Erickson et al., 2001; Erickson et al., 2003b; Pitt et al., 2001). Second, CaM can ‘bifurcate’ the local Ca²⁺ signal in the channel nanodomain: Ca²⁺ binding to the C-terminal lobe of CaM (C-lobe) alone can trigger one type of regulatory process on a channel; and Ca²⁺ binding to the N-terminal lobe (N-lobe) can induce a separate form of modulation (Chaudhuri et al., 2005; DeMaria et al., 2001; Yang et al., 2006). Intriguingly, despite evidence that a single CaM molecule orchestrates both processes (Mori et al., 2004; Yang, 2007), C-lobe induced regulation persists with millimolar concentrations of intracellular Ca²⁺ chelators like EGTA and BAPTA, whereas N-lobe processes are largely inhibited by such Ca²⁺ buffering (Liang et al., 2003). Not every channel manifests both C- and N-lobe functions, but the coexistence of both processes has been clearly demonstrated in Ca_V2.1 and Ca_V1.3 channels (Chaudhuri et al., 2005; DeMaria et al., 2001; Lee et al., 2003; Yang et al., 2006). Finally, the polarity of channel modulation by a given lobe of CaM can be inverted between closely similar Ca²⁺ channel types. For Ca_V2.1 channels, the C-lobe triggers a fast Ca²⁺-dependent facilitation of opening (CDF) (Chaudhuri et al., 2005; DeMaria et al., 2001), while the N-lobe produces a slower Ca²⁺-dependent inactivation (CDI) (Chaudhuri et al., 2005; DeMaria et al., 2001; Lee et al., 2003). By contrast, for Ca_V1.2 channels, Ca2+ binding to the C-lobe triggers a rapid CDI process, while the N-lobe initiates a more gradual and distinct CDI mechanism (Alseikhan et al., 2002; Dick et al., 2007; Peterson et al., 1999). Remarkably, structural similarities and single-channel data hint that the Ca_V2.1 and Ca_V1.2 C-lobe processes may be the same, only with opposing polarity (CDF versus CDI) (Chaudhuri et al., 2007). Probing this C-lobe directional inversion thereby promises rich mechanistic insights.

By contrast to function, less is known about the underlying structural bases ofCaM/channel regulation. Most of the structural determinants of CaM regulation of Ca_V1–2 channels are localized to their main, pore-forming α₁ subunits, specifically in the upstream third of subunit carboxy termini (Figure 1A, left, ‘CI’ region) (Evans and Zamponi, 2006). FRET and biochemical evidence indicates that apoCaM is preassociated with a consensus IQ domain, and an immediately upstream preIQ region (Erickson et al., 2001; Erickson et al., 2003a; Pitt et al., 2001). Ca²⁺ binding to CaM may cause rearrangements of CaM on these same segments (Evans et al., 2004), as Ca²⁺/CaM binds to IQ domains of many Ca_V1–2 channels (DeMaria et al., 2001; Peterson et al., 1999; Zuhlke et al., 1999), as well as to preIQ regions of Ca_V2.1 (Evans et al., 2004) and Ca_V1.2 (Erickson et al., 2003a; Kim et al., 2004; Pate et al., 2000; Pitt et al., 2001; Romanin et al., 2000; Tang et al., 2003). These rearrangements somehow modulate channel gating, with the potential involvement of an EF-hand-like channel module as a transduction element for C-lobe mediated processes (Chaudhuri et al., 2004; de Leon et al., 1995; Kim et al., 2004; Peterson et al., 2000; Zuhlke and Reuter, 1998). Given the multiplicity of structural determinants within a single molecular complex, the precise sequence underlying Ca²⁺-dependent regulation of channels remains unclear. Nonetheless, Ca²⁺/CaM interaction with the IQ domain has been viewed as potentially central to initiating a chain of ensuing regulatory events, because mutations within this domain can eliminate both CDF and CDI in Ca_V2.1 (DeMaria et al., 2001; Lee et al., 2003); and CDI within Ca_V1.2, Ca_V1.3, Ca_V2.2, and Ca_V2.3 channels (Evans and Zamponi, 2006; Yang et al., 2006). Indeed, these complexes between CaM and channel IQ domains represent sophisticated additions to a family of CaM—IQ assemblies, originally defined for CaM interactions with unconventional myosins, and now encompassing over 50 different mammalian molecules (Hoeflich and Ikura, 2002). Unlike classical CaM binding targets, IQ domains support both apoCaM and Ca²⁺/CaM interactions that underlie exquisite functional modulation of affiliated molecules (Black et al., 2006), nowhere more evident than in the setting of Ca²⁺ channels.

Figure 1. Structures of Ca²⁺/CaM bound to the IQ domain of Ca_V2.1 and Ca_V2.3.

(A) IQ domains of Ca_V Ca2+ channels. Left, cartoon of pore-forming α₁ subunits of Ca_V channels, with carboxytail landmarks, as labeled. EF, EF-hand motif; IQ, IQ domain; preIQ, intervening sequence important for apoCaM binding. CI region, structural ‘hot spot’ for CaM/channel regulation. Right, aminoacid sequence alignment of IQ domains of Ca_V channels, with IQ consensus pattern in bottom row. Upper calibration references amino acid coordinates for Ca_V2.1 α_1A subunit; bottom calibration gives universal coordinates with key isoleucine set to position 0. Yellow highlights, differences between Ca_V2.1 and Ca_V1.2 proposed to explain contrasting regulatory profiles between these two channel types. (B) Ribbon-diagram representations of Ca²⁺/CaM bound to the IQ domain of Ca_V2.3, from saggital (top) and coronal (bottom) viewpoints. The IQ domain is colored in red; amino-terminal end towards the left. N-and C-lobes are differentially colored as labeled. Side chain of I[0] residue shown as stick diagram. The four EF hands with Ca²⁺ (balls) are numbered in order. (C) Analogous Ca_V2.1 structure. (D) Superimposition of structures in panels B and C, indicating close similarity.

Very recently, the first atomic resolution structures of Ca²⁺ regulatory modules have been reported (Fallon et al., 2005; Van Petegem et al., 2005). In keeping with the presumed centrality of the IQ domain, these structures have focused upon Ca²⁺/CaM complexed with the IQ element of Ca_V1.2. Ca2+/CaM interacts extensively with this IQ domain, consistent with a Ca2+/CaM effector role for this site (Van Petegem et al., 2005). Interestingly, compared with Ca_V1 IQ domains, those for Ca_V2 channels differ at more than five positions (Figure 1A, right, yellow highlights), and these dissimilarities have been suggested by Van Petegem and colleagues to alter the structure of the IQ/CaM assembly, thereby accounting for the contrasting regulatory profiles of Ca_V2.1 versus Ca_V1.2 (Van Petegem et al., 2005). Given the rich mechanistic dividends pertaining to these oppposing modulatory profiles, this study explicitly tests for specific underlying differences in the structure of Ca_V2.1 versus Ca_V1.2 IQ domains in complex with Ca²⁺/CaM. Two crystal structures were here obtained: one for Ca²⁺/CaM in complex with the IQ domain of Ca_V2.3 channels; and the second for the analogous complex involving the IQ moiety of Ca_V2.1. Both structures were highly similar representatives of a generic Ca_V2 class structure, and correlation with functional analysis of an alanine mutational scan suggests that the C-lobe of CaM partially dislodges from the IQ element during channel regulation.

RESULTS

Recapitulation of distinctive features of the Ca_V IQ domains complexed with Ca²⁺/CaM

We utilized molecular replacement to solve two new crystal structures: one for Ca²⁺/CaM in complex with the IQ domain of Ca_V2.3 channels, resolved at a 2.2 Å resolution; and the second for the analogous complex involving the IQ moiety of Ca_V2.1 channels, at a 2.6 Å resolution. The search model was the analogous structure for Ca_V1.2 channels (C conformation, structure 2BE6) (Van Petegem et al., 2005). Ca_V2.3 and Ca_V2.1 complexes turned out to be highly similar, so the higher-resolution Ca_V2.3 structure best represents a generic Ca_V2 configuration. To facilitate comparison between IQ domains from different channels (Figure 1A, right), residues are numbered with the central isoleucine at position ‘0,’ here and throughout.

For the structure of Ca²⁺/CaM in complex with the Ca_V2.3 IQ domain, crystals exhibited an I4₁22 symmetry, and the asymmetric unit contained a single CaM/IQ complex with a 1:1 stoichiometry. The structure was notable in several regards (Figure 1B). First, CaM, as fully charged with four Ca²⁺ ions, envelopes an alpha helical IQ domain, which extends straight out without the mid-point bend seen in analogous Ca_V1 structures (Fallon et al., 2005; Van Petegem et al., 2005). Second, the Ca_V2.3 structure is also distinguished by a parallel orientation, wherein the N-lobe is positioned somewhat closer to the IQ amino terminus than is the C-lobe. This orientation is thus far unique to CaM/IQ structures of Ca_V channels (Fallon et al., 2005; Van Petegem et al., 2005), and CaM in complex with CaM-dependent kinase kinase (Osawa et al., 1999). Third, the two lobes of CaM are arrayed on opposite faces of the IQ helix; and the C- and N-lobes make little inter-domain contact compared to the majority of CaM/peptide structures (Hoeflich and Ikura, 2002), with only two residues from the different lobes of our structure residing within 4 Å of each other (T37 and N111). The features described under this third point enable CaM/IQ contacts to segregate between C- and N-lobe classes, as explicitly shown by the 4-Å contact map in Figure 2A. This segregation may facilitate the semi-autonomous regulatory action of these lobes in Ca_V1–2 channels (DeMaria et al., 2001; Lee et al., 2003; Liang et al., 2003; Peterson et al., 1999; Yang et al., 2006). Finally, the specifics of CaM interaction are characteristic of Ca²⁺/CaM—IQ complexes (Fallon et al., 2005; Van Petegem et al., 2005). Both C- and N-lobes made extensive hydrophobic and electrostatic contacts with the IQ segment (Figure 2A and 2C), respectively burying 582 and 702 Ų in surface area at the IQ domain interface. Additionally, not only were IQ residues traditionally associated with CaM interaction ([I/L/V]QxxxRxxxx[R/K]) found to be important (Jurado et al., 1999), but so were numerous other hydrophobic anchors distributed throughout the segment (Figures 2A and 2C). In particular, multiple contacts are present upstream of the I[0] position, and the total buried surface area of the IQ domain was evenly split between upstream sites (618 Ų), and those from the I[0] position onwards (653 Ų). These extensive upstream contacts contrast with the traditional emphasis on downstream consensus residues. Figure 2C displays specific examples of intimate docking at both upstream and downstream residues, emphasizing the fit of I[−6] within a hydrophobic pocket of the N-lobe (left), as well as the anchoring of Y[+3] and Y[+4] residues into like features of the C-lobe. Indeed, the IQ domain demonstrates greater solvent exposure in its downstream half, compared to an almost completely buried upstream portion (Figures 2A, bottom; Figure 2D). These features contrast with the dominance of C-lobe and downstream-IQ interactions within the apoCaM/IQ structure for myosin (Houdusse et al., 2006).

Figure 2. Residue-by-residue interaction of Ca_V2 IQ domains with Ca2+/CaM.

(A) Contact map of Ca²⁺/CaM with the IQ domain of Ca_V2.3. IQ domain sequence in light gray rectangle. CaM residues within 4 Å of indicated IQ domain position are shown above (N-lobe) and below (C-lobe) IQ sequence. Listings for CaM include number of atoms within each residue that are within 4 Å of the indicated IQ position. Water-mediated hydrogen bonds shown by ‘o’ symbol. Bar graph at the bottom plots solvent accessible area for each IQ position. (B) Contact map for Ca²⁺/CaM with Ca_V2.1 IQ domain of. (C) Close hydrophobic interactions between select IQ domain residues and hydrophobic pockets of Ca²⁺/CaM, taken from Ca_V2.3 structure (Figure 1B). Left, upstream residues -5 and -6 (shown as stick diagrams) dock closely with N-lobe pocket (shown in space-filling representation). Right, those at positions 3 and 4 nestle tightly with C-lobe cavities. CaM is color-coded for electrostatic potential (gray, hydrophobic; red, electronegative). (D) Space-filling representation of Ca_V2.3 structure. Partial exposure of IQ domain (red) at carboxy-terminal end. Y[+4], stick figure; V[+12], colored in yellow.

Crystals for the analogous Ca_V2.1 structure manifested a C2 symmetry, where the asymmetric unit contained two CaM/IQ complexes (1 and 2), each with 1:1 stoichiometry. Both complexes adopted nearly indistinguishable conformations, though complex 1 yielded a somewhat more defined electron density map. Accordingly, all subsequent analysis refers to structure 1 (Figures 1C and 2B). The near identity of this Ca_V2.1 structure to its Ca_V2.3 counterpart is documented by the overlay in Figure 1D. Additionally, the detailed contact map (Figure 2B) confirmed a highly analogous interface between CaM and the IQ domain of Ca_V2.1, and the atomic positions were closely similar throughout (Table 1). The details of model refinement for both structures are summarized in Table 2.

Table 1.

Cα-RMS deviations (Ų) for structures of Ca²⁺/CaM complexed with IQ domains of Ca_V2.3, Ca_V2.1, and Ca_V1.2

	CaV2.3	CaV2.1a	CaV2.1b	CaV1.2*	CaV1.2#
CaV2.3		0.93	0.70	0.99	2.51
CaV2.1a			0.55	1.01	2.40
CaV2.1b				0.90	2.42
CaV1.2*					2.37
CaV1.2#

The superscripts ‘a’ and ‘b’ denote structures 1 and 2 of the Ca_V2.1 complex, respectively.

^*C-form in 2BE6.pdb.

^#2F3Y.pdb.

Table 2.

Data collection and refinement statistics

	CaV2.3 IQ/CaM	CaV2.1 IQ/CaM
Crystal parameters
Space group	I4122	C2
Cell constants (Å)	124.61, 124.61, 74.02	80.66, 80.43, 63.84
Data collection
Wavelength (Å)	1.54	1.54
Resolution (Å)	30-2.2 (2.28-2.2)	50-2.6 (2.7-2.6)
Rsym	0.089 (0.502)	0.061 (0.35)
I/σI	17.3 (1.8)	6.9 (2.1)
Completeness (%)	98.57 (92.7)	92.39 (94.5)
Redundancy	6.3 (5.1)	1.6 (1.6)
Refinement
Resolution (Å)	30-2.3	24.9-2.6
No. reflections	12496	12283
Rwork/Rfree	20.5/26.1	20.8/25.3
No. atoms
Protein	1291	2523
Ligand/ion	14	23
Water	138	53
B-factors
Protein	40.30	68.86
Ligand/ion	67.92	67.02
Water	46.87	70.32
RMS deviations
Bond lengths (Å)	0.008	0.006
Bond angles (°)	1.009	0.849

Highest resolution shell shown in parentheses.

Thus oriented to Ca_V2 structure, we searched for any structural differences relative to the analogous complex for Ca_V1.2 channels, ones that might rationalize the notable contrasts in CaM regulatory phenotype between these two channel classes. In this regard, no simple answers arose: the Ca_V2 structure is highly similar to the most common Ca_V1 form, but for the lack of an ~10° bend in the mid IQ segment of the Ca_V1 structure (Supplemental Data 1). This minor difference, in a largely buried peptide, seemed insufficient to explain appreciable functional differences. Also, the similarity of an extensive contact map for a published Ca_V1.2 complex (Van Petegem et al., 2005) (Supplemental Data 1), and statistical comparisons of atom positions (Table 1), underscored the overall structural similarities.

Systematic alanine scanning of the Ca_V2.1 IQ domain

To gain perspective on the contrast between structural similarity and functional divergence, we undertook alanine-scanning mutagenesis of the Ca_V2.1 IQ domain, and correlated mutations with CaM regulatory function as assayed by whole-cell electrophysiological recordings. Previous reports have described the effects of cluster mutations within the IQ domain (DeMaria et al., 2001; Lee et al., 2003), but these scans were far from exhaustive, and could well involve perturbations of main-chain conformation that complicate interpretation. By contrast, point alanine mutations probe side-chain effects, without perturbation of the backbone fold (Cunningham and Wells, 1989).

Figure 3A introduces a first form of CaM regulation in Ca_V2.1: Ca2+ binding to the C-lobe of CaM triggers a Ca²⁺-dependent facilitation (CDF) of current that accrues over tens of milliseconds, before appreciable CDI becomes apparent (Chaudhuri et al., 2005; Chaudhuri et al., 2007; DeMaria et al., 2001; Soong et al., 2002). With Ca²⁺ as charge carrier, CDF is readily detected as a slow phase of increasing Ca²⁺ current seen during short (50-ms) test depolarizations (top, gray trace, arrow). When channels are facilitated by Ca²⁺ entry during a preceding voltage prepulse, ensuing currents activate rapidly to the facilitated level during the test-pulse (top, black trace). To quantify the facilitation produced by the prepulse, we integrate the difference between normalized test-pulse currents in the absence and presence of a prepulse (ΔQ, gray shaded area), and this integral is used to determine the relative facilitation (RF) induced by the voltage prepulse. On average, RF demonstrates a bell-shaped dependence upon prepulse voltage (bottom, open circles), as expected for a genuine Ca²⁺-driven process (Brehm and Eckert, 1978). By contrast, fitting with the strong preference of CaM for Ca²⁺ over Ba²⁺ (Chao et al., 1984), there is little evidence of such prepulse facilitation with Ba²⁺ as charge carrier (bottom, filled circles). The small rise of RF seen with Ba²⁺ (~0.05, bottom, open circles) likely reflects background G-protein modulation (DeMaria et al., 2001). To index pure CDF, we consider the increase of the Ca²⁺ over Ba²⁺ RF relations (g ~0.2), following a 20-mV prepulse.

Figure 3. Functional alanine scan of Ca_V2.1 IQ domain.

(A) Baseline CDF for wild-type Ca_V2.1 channel. Top, exemplar Ca²⁺ current traces evoked by a test pulse, with (black) and without (gray) a voltage prepulse. Dark gray shading emphasizes the facilitation produced by Ca²⁺ entry during the prepulse. Bottom, population data for CDF. RF, relative facilitation produced by voltage prepulses to the potential shown on the abscissa. Open symbols, data with Ca²⁺ currents; filled symbols, data with Ba²⁺ currents. Difference between relations at +10 mV (g) quantifies pure CDF. Numbers in parentheses, number of cells in population data, here and throughout. (B, C) Elimination of CDF for corresponding I[−6]A (B) and M[−1]A (C) mutations. (D) Baseline CDI profile for wild-type Ca_V2.1 channel. Top, exemplar Ca²⁺ current (gray) and Ba²⁺ current (black), as evoked by 10-mV test pulse. Light gray shading emphasizes extent of CDI. Current scale bar pertains to Ca²⁺ current. Ba²⁺ current scaled down 2–3× to facilitate visual comparison of decay kinetics. Bottom, population data for CDI. r₈₀₀ is the fraction of peak current remaining after 800-ms depolarization to the potential shown on the abscissa. Open symbols, for Ca²⁺ currents; filled symbols, for Ba²⁺ currents. Difference between relations at +10 mV (f) quantifies pure CDI. (E) CDF of Ca_V2.1 mutants, with changes to alanine as indicated. Bottom rows indicate predominate CaM lobe contact (C, N), and Robetta calculated disruption of binding energy upon mutation to alanine. (E) CDI of Ca_V2.1 mutants, with changes to alanine as indicated. (G) Electron density contour around Ca_V2.3 IQ domain from positions −7 to +9. Mesh encloses region where density exceeds background by two standard deviations, as generated by CCP4 mg.

Given this baseline, we examined the effects of alanine point mutations across the entire Ca_V2.1 IQ domain. To start, we recalled that a double alanine mutation in the central IQ domain residues IM→AA (positions 0–1) eliminated CDF in prior work (DeMaria et al., 2001). Dissecting this result into single point mutations (isoleucine at position zero mutated to an alanine (I[0]A); and methionine at position 1 changed to alanine (M[1]A)) indicates that the bulk of attenuation is attributable to the I[0] position (Figure 3E). Beyond this derivative result, the results of alanine scanning were unexpected (Figure 3E). Point mutations at any of the other canonical IQ domain residues (gray highlighted residues) had no appreciable effect on CDF. In fact, no more than minor effects were observed for any of the residues downstream of the initial I[0] at the beginning of the canonical motif; the strongest downstream effect was seen with Y[+3]A and Y[+4]A perturbations, which induced no more than 30–40% attenuation of function. The conventional concept for IQ domains predicts that the region downstream of I[0] would be functionally dominant (Bahler and Rhoads, 2002). By contrast, alanine point mutations at multiple upstream sites induced surprisingly large reductions in CDF; and this upstream trend was upheld for one of two naturally occurring alanines that were changed to threonines (Figure 3E, positions −4 and −3). In particular, CDF was nearly abolished with I[−6]A, A[−4]T, and M[−1]A mutations (Figures 3B, 3C, and 3E), all of these being upstream residues. These findings emphasized a surprising structure-function correlation for the Ca²⁺-CaM/IQ complex of Ca_V2 channels: IQ domain contacts at multiple residues upstream of I[0] were functionally dominant for C-lobe triggered regulatory function, compared with the canonical IQ motif residues, which include I[0] and selected downstream positions. This feature likely relates to the substantial structural contrasts between Ca²⁺-CaM/IQ and apoCaM/IQ complexes (Houdusse et al., 2006), for which the latter forms the basis of traditional IQ motif hotspots (Figure 3E, highlighted positions). The set of key upstream residues identified here may be emblematic of a new canonical IQ motif pertinent to Ca²⁺/CaM interactions. In this regard, the pattern of upstream dominance may well pertain to the Ca_V1 Ca²⁺-CaM/IQ complex in regard to C-lobe signaling (Fallon et al., 2005; Van Petegem et al., 2005).

We next turned to a second form of CaM regulation of Ca_V2.1 channels, as shown in Figure 3D. Ca²⁺ binding to the N-lobe of CaM triggers a Ca²⁺-dependent inactivation (CDI) that progresses over hundreds of milliseconds (Chaudhuri et al., 2005; Chaudhuri et al., 2007; DeMaria et al., 2001; Lee et al., 2003; Soong et al., 2002). This can be seen from the faster decay of Ca²⁺ versus Ba²⁺ currents (Figure 3D, top). Because the Ba²⁺ current decay reflects a separate form of voltage-dependent channel inactivation (Alseikhan et al., 2002; Jones et al., 1999; Patil et al., 1998), the additional speeding of decay seen in Ca²⁺ versus Ba²⁺ currents reflects pure CDI (shaded area). On average, inactivation can be thus quantified by plotting the fraction of peak current remaining after 800-msec depolarization to various voltages (r₈₀₀, Figure 3D, bottom). The maximum difference between Ba²⁺ and Ca²⁺ relations, f₈₀₀ = 0.36, quantifies average CDI. Whereas substitution of glutamate at IQ positions 0 and 1 does eliminate this CDI (DeMaria et al., 2001), more targeted alanine substitutions have left CDI unchanged, at least for the few residues where such alanine substitutions have thus far been made (DeMaria et al., 2001; Lee et al., 2003). Here, despite an exhaustive scan, none of the alanine cluster/point mutations throughout the IQ domain appreciably affected CDI (Figure 3F). This result was unexpected, given the anticipated centrality of the IQ domain for CaM/channel regulation.

Potential mechanisms of CaM/channel regulation implicated by Ca²⁺-CaM/IQ structures

The demonstrated similarity in Ca²⁺-CaM/IQ structure for Ca_V2 and Ca_V1 channels (Table 1; Supplemental Data 1), together with the functional outcomes of systematic alanine scanning of the Ca_V2.1 IQ region (Figure 3), offered new constraints for evaluating mechanisms of CaM/channel regulation. The simplest scenario for rationalizing the opposing regulatory polarities of C-lobe regulation in Ca_V2.1 and Ca_V1.2 channels would arise from clear distinctions in the Ca²⁺-CaM/IQ structures of Ca_V2.1 versus Ca_V1.2 channels. Our new structures (Figures 1 and 2) clearly exclude this scenario, at least for the conformation in our crystals. This exclusion allowed us to focus on two other broad classes of mechanisms.

Class 1

This mechanism postulates insignificant structural differences between Ca_V2.1 and Ca_V1.2 channels, when comparing either their apoCaM/IQ complexes, or their Ca²⁺-CaM/IQ assemblies. In this manner, the IQ domain itself would play no distinguishing role in specifying C-lobe signaling polarity. Instead, nearly identical Ca²⁺-induced conformational changes in the CaM/IQ structure would be decoded in different ways by other parts of these two channel types. An important corollary of this class is ‘context-independent IQ function,’ wherein substitution of the IQ domain from one channel subtype into another should not alter C-lobe signaling.

To test this prediction, we undertook the chimeric channel experiments displayed in Figure 4. Ca_V1.2 and Ca_V2.1 channels (with pore-forming subunits α_1C and α_1A, respectively) both manifest Ca²⁺ regulation triggered by Ca²⁺ binding to the C-lobe of CaM, but the regulatory polarity is CDI for Ca_V1.2, and CDF for Ca_V2.1. To examine the effects of targeted switching of IQ domains, we substituted the crucial proximal third of the carboxy terminus (Figure 1A, CI region) from either an α_1C or α_1A subunit onto a single, expression-permissive Ca_V2.3 channel backbone (Figure 4A, left, α_1E). Direct swapping of CI regions between α_1C and α_1A subunits yielded nonfunctional channels (not shown), while Ca_V2.3 channels (α_1E subunit backbone) do not exhibit intrinsic C-lobe CaM regulation (Liang et al., 2003). Hence, the Ca_V2.3 backbone furnished an appropriate and uniform context in which the regulatory effects of chimeric channel manipulations could be evaluated. Importantly, all recordings were performed in elevated Ca²⁺ buffering (5 mM EGTA) so as to isolate the C-lobe component of CaM regulation, and the electrophysiological protocols and analysis were analogous to those in Figures 3A and 3D.

Figure 4. Context-dependent IQ function.

(A) C-lobe regulation in chimeric channel with carboxy tail from Ca_V1.2 α_1C subunit (construct 1). Left, cartoon of chimeric channel α₁ subunit composition, with carboxytail landmarks as in Figure 1A. The ‘backbone’ of all chimeric α₁ subunits, from the amino terminus through the beginning of the carboxy terminus, is derived from the Ca_V2.3 α_1E subunit, here and throughout the figure. The proximal third of the carboxy tail from the Ca_V1.2 α_1C subunit, spanning the EF-hand through the IQ domain, is fused to this backbone to create construct 1. Middle, exemplar Ca²⁺ traces showing absence of CDF (format as in Figure 3A). Right, exemplar Ca²⁺ and Ba²⁺ traces illustrating the presence of CDI (format as in Figure 3D). All experiments utilized high 5 mM EGTA Ca²⁺ buffering, to enrich for C-lobe mediated regulation throughout the figure. (B) C-lobe regulation in chimeric channels with carboxy tail from Ca_V2.1 α_1A subunit (construct 2). CDF is present, but CDI is absent. (C) CDF and CDI population data, for constructs 1–6. Left, schematic showing composition of constructs. Middle, average CDF strength (g). Right average CDI strength (f). Metrics defined in Figures 3A and 3D. Context-dependent function of IQ domain shown by constructs 2 versus 3, and constructs 5 versus 6.

The experimental results clearly contradicted context-independent function of the IQ domain. For construct 1, with the carboxy tail from Ca_V1.2 channels, C-lobe signaling induced CDI (Figures 4A and 4C, top row), as for the parent Ca_V1.2 channel. Following the same linkage to the parental channel, construct 2 (carboxy tail from Ca_V2.1 channels) exhibited CDF (Figures 4B and 4C, row 2). Given this baseline, mechanistically informative results were obtained by selective exchange of IQ segments between constructs 1 and 2. Figure 4C (constructs 3 and 4) shows that this exchange did not preserve the C-lobe signaling present in the original constructs. Construct 3, which features insertion of the IQ segment of Ca_V2.1 into construct 1, failed to exhibit appreciable CDF or CDI. Construct 4, wherein the IQ segment from Ca_V1.2 has been introduced into construct 2, shows significantly weaker CDF than seen in construct 2. Yet stronger violation of context-independent IQ function came from two further constructs. Upon substituting the EF-hand domain from Ca_V1.2 into construct 2, the resulting construct 5 still exhibited CDF (Figure 4C, row 5), albeit weaker than in construct 2. However, after switching the Ca_V1.2 IQ domain into construct 5, the resulting construct 6 showed unmistakable CDI (Figure 4C, row 6). Thus, substitution of an IQ domain alone, as seen in the conversion from construct 5 into 6, switched regulatory polarity. Hence, class 1 cannot be true.

Class 2

This leads to a remaining possibility, wherein our Ca_V2 IQ/CaM structure might only be partially related to the conformation underlying functional CDF. Instead, the physiologically relevant conformation features some displacement of CaM to expose portions of the IQ domain for interaction with other regions, which in turn triggers channel regulation. In this scenario, the IQ domain would contribute in part to C-lobe signaling polarity, while interactions with other portions of the channel would also matter.

To investigate this hypothesis, we compared our Ca_V2 IQ/CaM structure (Figure 1) to the functional outcomes from alanine scanning mutagenesis (Figure 3) using Robetta, a molecular simulation package for ligand/receptor binding energy perturbation via single point mutations (Kortemme and Baker, 2002). These algorithms have successfully predicted these energies for multiple sets of experimentally determined structures and binding energies. Here, we utilized our Ca_V2.3 IQ/CaM structure as an input to Robetta, and the energetic effects of various point mutations to binding were then estimated. Importantly, a core assumption of this algorithm, that main chain conformation undergoes little perturbation by mutations, is consistent with prior structures showing that Ca²⁺/CaM in complex with a mutant Ca_V1.2 IQ domain is nearly indistinguishable from the wild-type configuration (Fallon et al., 2005). Figure 3E (bottom) displays the results of the Ca_V2.3 analysis as ΔΔG values, each denoting the estimated change in IQ/CaM binding energy for the indicated point mutation. All portions of the IQ segment subject to these energy calculations were characterized by well-defined electron densities, as demonstrated in Figure 3G (bottom). If the extent of IQ/CaM binding, as captured in our crystal structure (Figure 1B), were directly related to the strength of functional CDF, then plotting the strength of CDF (g) as a function of ΔΔG should define a 1:1 binding isotherm. Alternatively, if such binding were only indirectly related to function, such as in a hypothetical configuration featuring some Ca²⁺/CaM displacement towards another surrogate partner, an absence of correlation between these two variables might result. Figure 5A displays the results of this analysis, where each symbol relates to a single point mutation from Figure 3E. Initial examination would suggest no correlation. However, when the data were segregated according N- and C-lobe interfaces, a coherent pattern emerged. The plot for data relating to the N-lobe interface alone (Figure 5B) renders a clearly recognizable 1:1 binding curve, with half saturating binding energy ~2 kcal/mole. To further corroborate this relation, we introduced mutations other than alanine at a single position (I[−6]), where intimate CaM/IQ contact is prominent in Ca_V2 structure (Figure 2C). Reassuringly, these alternative mutations produced data that also resided close to the same binding curve (Figure 5B, open symbols). Hence, IQ domain binding to the N-lobe of CaM appears intimately related to functional CDF.

Figure 5. Comparing CDF function and predicted binding energy perturbation.

(A) Apparent absence of correlation between strength of CDF (ordinate) and predicted CaM/IQ binding perturbation by point mutations in IQ domain of Ca_V2 channels (abscissa). Filled symbols correspond to data from Figure 3E. (B) Strength of CDF correlates well with predicted binding energy perturbations, if analysis is restricted to IQ residues with N-lobe contacts. Filled symbols, N-lobe subset of data in panel A; open symbols, additional mutations at position −6. Solid curve plots 1:1 binding isotherm, with a half-binding energy value of ~2 kcal/mol. (C) Discordance of strength of CDF and predicted binding energy perturbations for C-lobe contacts. Filled symbols, C-lobe subset of data in panel A; open symbols, additional mutations at positions −1, +3, and +4. Locus 1, cases where energy calculations predict major disruption of CaM binding to IQ domain, yet CDF is largely preserved. Locus 2, cases where energy predictions suggest little perturbation of CaM binding to IQ domain, yet there is nearly complete elimination of CDF. (D) Y[+3]E mutant preserves CDF. (E) Y[+4]E mutant also spares CDF.

By contrast, the analogous plot for the C-lobe interface retained a disordered pattern (Figure 5C). In particular, two notable groups of discordance were present. First, Y[+3] and Y[+4] residues both anchor the C-lobe of CaM in our structure (Figure 2C), and corresponding Y[+3]A and Y[+4]A mutations are each predicted to markedly attenuate IQ/C-lobe binding (ΔΔG ~3 kcal/mole). Yet, the effect on CDF was minimal (Figure 5C, locus 1). Second, K[−7], A[−4]T, and M[−1] mutations should produce minimal perturbation of the IQ/C-lobe interface (locus 2, ΔΔG ~1 kcal/mole or less), yet extremely strong reductions of CDF were produced. More extreme forms of such divergence came from less conserved mutations at two of these key sites (Figure 5C, open symbols). For locus 1, both Y[+3]E and Y[+4]E mutants project an enormous ΔΔG ~6 kcal/mole, yet CDF is still largely maintained at the levels for alanine mutations (Figures 5C, 5D and 5E). For locus 2, the M[−1]T mutation should barely perturb IQ/C-lobe interaction, yet there is still marked diminution of CDF (Figures 5C).

These examples of C-lobe discordance suggest that the actual CDF configuration involves some displacement of the C-lobe away from the IQ domain (i.e., class 2), towards an alternate binding site that minimizes overall energetics. Nearby Ca²⁺/CaM binding sites, just upstream of the IQ domain, have been identified in Ca_V2.1 (Evans et al., 2004), and in Ca_V1.2 (Erickson et al., 2003a; Kim et al., 2004; Pate et al., 2000; Pitt et al., 2001; Romanin et al., 2000; Tang et al., 2003). Also, the presumed departure of the C-lobe from the IQ domain might be facilitated by heightened solvent accessibility of the IQ domain downstream of I[0] (Figure 2A, bottom; 2B, bottom; 2D), given that the C-lobe of CaM interacts to a somewhat greater degree with this downstream IQ region. Additionally, prior biochemical experiments demonstrate that cysteine substitution at position 0 of the Ca_V1.2 IQ domain yields Ca²⁺-CaM/IQ dimers linked by disulfide bonds (Kim et al., 2004). Finally, partial C-lobe departure from the IQ element would render CaM/IQ conformations distinguishable between Ca_V1.2 and Ca_V2.1 contexts; in particular, customized interaction of an exposed IQ surface with other channel domains explains both the divergence of C-lobe regulatory polarity between Ca_V1.2 and Ca_V2.1 channel subtypes, as well as the context-dependent functionality of the IQ domain in Figure 4. Cursory examination of Figure 4C might even suggest that it is complementation between EF and exposed IQ surfaces that specifies C-lobe polarity. However, the lack of CDF and CDI in an ‘ACCA’ chimeric channel (not shown) suggests that such IQ interactions might extend beyond the EF element alone. Previously reported data hint at potential extensions to the cytoplasmic loop between domains I and II of Ca_V1.2 channels (Kim et al., 2004). Overall, the weight of evidence, including our energy calculations of Ca_V2.1 point mutations, favors some C-lobe displacement from the IQ domain during channel regulation (class 2). A like scenario has been suggested for Ca_V1.2 (Fallon et al., 2005), but here we explored this proposal in greater depth.

DISCUSSION

We have solved the crystal structure for Ca²⁺/CaM in complex with the IQ domains from Ca_V2.3 and Ca_V2.1 channels. Both of these may be considered as equivalent to a generic Ca_V2 structure, which shares most of the hallmark features for previously resolved structures of Ca²⁺/CaM bound to the IQ domain of Ca_V1.2 channels. One surprising aspect of this structure arises by alanine scanning mutagenesis of the entire Ca_V2.1 IQ domain. For regulation by the C-lobe of CaM, IQ domain contacts upstream of the canonical IQ motif (IQxxxRGxxxR) are functionally dominant, while the downstream canonical contacts play a lesser role. This set of key upstream residues encompasses hydrophobic contacts at [−6], [−5] and [−1] positions. As these are not predicted by any known Ca²⁺/CaM binding motif (Rhoads and Friedberg, 1997), this set of contacts may comprise part of a new canonical IQ motif pertinent to Ca²⁺/CaM interactions (as opposed to the traditional motif for apoCaM (Rhoads and Friedberg, 1997)). Fitting with potential generalization, an IQ-domain data base indicates conservation of such hydrophobic residues (Marchler-Bauer et al., 2007). Another interesting feature concerns signaling triggered by Ca²⁺ binding to the N-lobe of CaM (Figure 3D). In this context, none of the IQ domain contacts appeared functionally consequential. A final key result is that only minor structural differences are present between the Ca_V2 and Ca_V1 IQ/CaM structures, and these in themselves seem insufficient to account for the substantial differences in C-lobe CaM-mediated regulation of these channels. Together, these results unambiguously exclude the simplest of prevailing hypotheses for channel regulation, wherein clear distinctions in the Ca²⁺-CaM/IQ structures of Ca_V2.1 versus Ca_V1.2 channels are primarily responsible for divergent regulatory effects. This exclusion now emphasizes more complex mechanisms, wherein multiple binding sites and structural determinants co-dominate in specifying CaM regulation of channels.

Specifically, by using molecular simulation (Robetta) to correspond our Ca_V2 structure with the functional outcomes of alanine scanning mutagenesis across the IQ domain of Ca_V2.1 (Kortemme and Baker, 2002), we favor a working proposal (class 2) in which the configuration in our Ca_V2 Ca²⁺-CaM/IQ structure may not fully capture the configuration underlying functional CDF. Instead, the C-lobe may partially dislodge from the IQ segment, enabling contacts between exposed IQ surfaces and other channel domains to trigger C-lobe mediated channel regulation. These alternative contacts may be customized to produce opposing C-lobe regulatory effects between Ca_V2.1 and Ca_V1.2 channels. Likewise, the absence of appropriate contacts may explain the lack of C-lobe mediated regulation in Ca_V2.2 and Ca_V2.3 channels (Liang et al., 2003). Importantly, this proposal rationalizes the overall sparing of C-lobe driven CDF by mutations at Y[+3] and Y[+4] positions (Figure 3E), despite intimate anchoring of these residues within a C-lobe cavity of the crystal structure (Figure 2C).

A remaining challenge for this class 2 proposal concerns the suggested importance of IQ contacts with the N-lobe in producing CDF (Figure 5B); this importance might initially appear to contradict previous results that Ca²⁺ binding only to the C-lobe of CaM is required to produce CDF (Chaudhuri et al., 2005; DeMaria et al., 2001). However, while our structure contains the Ca²⁺-bound N-lobe in complex with the IQ domain, the contacts and energetics for a Ca²⁺-free N-lobe interaction with the IQ element may be similar, and we would propose that either form of N-lobe/IQ binding supports CDF. Another form of Ca²⁺/CaM displacement could explain the uniform absence of IQ mutational effects on regulation triggered by Ca²⁺ binding to the N-lobe of CaM (Figures 3D and 3F). In particular, displacement of the N-lobe from the IQ domain, and subsequent binding to another site, could trigger the slower form of CaM modulation (CDI, Figure 3D), in a manner insensitive to IQ manipulation. Indeed, there is evidence that the Ca²⁺-bound N-lobe interacts with an amino-terminal channel locus to produce CDI in Ca_V1.3 and Ca_V1.2 channels (Dick et al., 2007), and we would propose that Ca_V2 channels contain a similar site elsewhere in the channel. Given this view, the slow induction of CDI would terminate CDF, because CDF appears reliant upon N-lobe/IQ interactions.

One other class of mechanism deserves consideration. There could be appreciable differences in the structures of apoCaM in complex with the IQ domains of Ca_V2.1 and Ca_V1.2 channels, while the analogous structures for Ca²⁺/CaM are very similar. This proposal readily explains differences in regulatory polarity, in terms of distinct apoCaM/IQ configurations for different channel types. Moreover, this scheme could partially explain the alanine scanning data for Ca_V2.1 (Figure 4C), as some discord between function and Ca2+/CaM binding to the IQ domain might reflect actual changes in apoCaM—IQ binding. However, live-cell FRET experiments between apoCaM and holochannels indicate a similar apoCaM/channel configuration in Ca_V1.2, Ca_V2.1, and Ca_V2.3 contexts (Erickson et al., 2001), at odds with this scenario. Though more direct data would be required to conclusively exclude this proposal, we favor class 2 mechanisms as more consistent with a comprehensive view of experimental constraints.

In all, our results emphasize that CaM/channel regulation likely reflects a multiplicity of binding configurations and structural determinants, likely extending beyond the IQ domain (Dunlap, 2007). Combined consideration of structure, alanine scanning, and molecular simulation hint at a scenario wherein Ca²⁺/CaM displaces from the IQ domain to facilitate further interactions triggering channel regulation. This provisional proposal forms an important framework for ongoing work. Future studies must peer beyond Ca²⁺/CaM in complex with the IQ domain, refine identification of additional effector sites, and resolve further related structures at the atomic level. In these structures, it may be necessary to simultaneously incorporate CaM with multiple sites, to fully capture the CaM/channel configurations most closely linked to modulation of channel gating. In this process, simple forms of molecular simulation (Baker, 2006) to correlate function with structure may continue to prove advantageous.

EXPERIMENTAL PROCEDURES

Protein purification and crystallization

Recombinant rat CaM (as cloned into pET24b, Novagen, EMD Chemicals, San Diego, CA) was expressed in E. coli BL-21(DE3), and purified to homogeneity by octyl-sepharose column chromatography (GE Healthcare Piscataway, NJ). The purified CaM was shown to be greater than 95% pure by SDS-PAGE. Synthetic peptides of the IQ domains for Ca_V2.1 and Ca_V2.3 channels were generated at the Synthesis and Sequencing Facility at the Johns Hopkins University School of Medicine. To confirm molecular weights, these peptides were analyzed by MALDI-TOF mass spectrophotometry (Applied Biosystems, Foster City, CA). Mixtures of CaM (10 mg/ml) and each IQ-domain peptide (3 mg/ml) were incubated in a solution containing 20 mM MOPS-NaOH, 10 mM CaCl₂, at pH 7.4. The mixture was then applied to a gel filtration column (Superdex 75 10/300, GE Healthcare), equilibrated with same buffer. A purified CaM/peptide complex was thus isolated, and subsequently crystallized using hanging drop vapor diffusion. Hanging drops were comprised of 1 μl of protein solution and 1 μl of a 1:2 dilution of reservoir buffer (described below) with water. Each drop was vapor equilibrated against 500 μl of undiluted reservoir buffer, and incubated at 18 °C. Further details of the crystallization conditions and related information for each type of complex are elaborated below.

Ca_V2.1 IQ-CaM complex—The purified complex of CaM and the IQ domain of Ca_V2.1 was further concentrated to 10 mg/ml, using a Centricon spin column (YM-3, Millipore, Bedford, MA). This concentrated complex was mixed into droplets as described above, with a reservoir buffer comprised of 2.3 M ammonium sulfate, 0.15 M sodium tartrate, and 0.05 M sodium-citrate (pH 5.5). Jagged, trapezoidal-shaped crystals appeared within four days, and achieved their full size in approximately two weeks. Diffraction quality crystals were grown by micro-seeding initial jagged crystals into freshly mixed droplets that had been pre-equilibrated for 3 hours. This produced single, rectangular-shaped crystals that grew in space group C2 and contained two CaM/IQ complexes in the asymmetric unit. The crystals used for x-ray diffraction analysis were soaked briefly in a fresh aliquot of reservoir buffer. The diffraction data were collected at room temperature, with crystals held in a wax-sealed glass capillary tube. The x-ray source was CuKα radiation produced by a Rigaku RU-200 generator (Tokyo, JAPAN).

Ca_V2.3 IQ-CaM complex—The purified complex of CaM and the IQ domain of Ca_V2.3 was concentrated to 10 mg/ml, mixed into droplets as above, and crystallized using a reservoir buffer containing 2.1 M ammonium sulfate, and 0.05 M sodium-citrate (pH 5.0). Pyramidal crystals appeared within two days, and achieved their full size in approximately 1 week. Crystals exhibited a I4₁22 space group, with one CaM/IQ complex within the asymmetric unit. The crystals used for diffraction analysis were soaked briefly in a cryoprotectant (20% glycerol in reservoir buffer), and flash frozen in a gaseous nitrogen stream at −180 °C. Diffraction data were then collected as above.

Analysis of x-ray diffraction data

Data were processed using HKL2000 (Otwinowski and Minor, 1997). The C-lobe of CaM, taken from a previously solved structure of Ca²⁺/CaM in complex with the IQ domain of Ca_V1.2 channels (PDB ID code 2BE6 C-form), was used as the search model in a molecular replacement approach to solving the structure of Ca²⁺/CaM in complex with the IQ domain of Ca_V2.3. In this process, we used the program MOLREP (Vagin and Teplyakov, 1997) from the CCP4 program suite (Collaborative Computational Project, 1994). Structure refinement was accomplished with iterative rounds of model building and refinement, using COOT (Emsley and Cowtan, 2004) and REFMAC5 (Murshudov et al., 1997) software packages, respectively. The structure for Ca²⁺/CaM in complex with the IQ domain of Ca_V2.1 was obtained in a similar manner, except that the search model was the newly solved structure for Ca²⁺/CaM in complex with the IQ domain of Ca_V2.3. In addition, the refinement also incorporated TLS algorithms, with TLS groups derived from TLSMD (Painter and Merritt, 2006). All data collection and model refinement statistics for both structures are summarized in Table 2. Molecular graphics were prepared with CCP4 mg software (Potterton et al., 2004). Total buried surface area was determined by the AREAIMOL program in the CCP4 suite (Lee and Richards, 1971), as applied to our Ca_V2.3 structure. Coordinates and structure factors will be deposited in the Protein Data Bank, www.pdb.org (accession codes XXXX and YYYY for Ca_V2.1 and Ca_V2.3, respectively).

Transfection of human embryonic kidney 293 cells

Human embryonic kidney (HEK293) cells were transiently transfected (calcium phosphate protocol) and cultured in 6-cm plates, as described (Brody et al., 1997). For expression of recombinant calcium channels, we applied 4 μg of cDNA encoding the desired α₁ subunit (Ca_V), along with 4 μg each of rat brain α_2a (Perez-Reyes et al., 1992), rat brain α_2δ subunits (Tomlinson et al., 1993), and finally 1 μg of the RSV T antigen. The α_1A pore-forming subunit (Ca_V2.1) (Soong et al., 2002) was cloned from a human source (NM000718), and the Ca_V1.2 α_1C subunit from rabbit (Wei et al., 1991) (X15539). All chimeric constructs were made from the carboxy tails of these two parental α_1A and α_1C channel subunits, as fused in various combinations onto the rat brain Ca_V2.3 α_1E subunit (Soong et al., 1993) (NM019294), which served as a channel backbone (Figure 4).

Electrophysiology

Whole-cell current recordings were performed 1–3 days after transfection, using glass pipettes with resistances of 1.5–3 MΩ prior to series resistance compensation. For characterization of regulatory processes initiated by Ca²⁺ binding to the C-lobe of CaM (all Figures except 3D and 3F), which can occur when Ca²⁺ elevations are restricted to within tens of nanometers of the channel cytoplasmic pore, the internal solution contained a relative high concentration of Ca²⁺ buffer, with a specific composition containing: Cs-MeSO₃, 135 mM; CsCl, 5 mM; EGTA, 5 mM; MgCl₂, 1 mM; MgATP, 4 mM; and HEPES (pH 7.3), 10 mM; 290 mOsm, adjusted with glucose. To characterize N-lobe mediated channel regulation, which favors global Ca²⁺ elevations, the EGTA was reduced to 1 mM in the internal solution (Figures 3D and 3F, only). The bath solution contained: TEA-MeSO₃, 140 mM; HEPES (pH 7.3), 10 mM; and CaCl₂ or BaCl₂, 5 mM; 300 mOsm, adjusted with glucose. Standard patch-clamp techniques were used with an Axopatch 200A amplifier (Axon Instruments/Molecular Devices, CA). Currents were filtered at 2 kHz and sampled at 10 kHz; series resistance was typically 1–2 MΩ after >70% compensation; and leaks and capacitive transients were subtracted by a P/8 protocol. Test pulse depolarizations were delivered every 30 s for local Ca²⁺ regulation protocols, and 60 s for global Ca²⁺ regulation protocols. Population data are presented as mean ± SEM, after analysis by custom-written software in MATLAB software (MathWorks, MA) and Microsoft Excel.

Construction of mutant and chimeric molecules

Single or cluster mutations in the IQ domain were generated using the QuikChange procedure (Stratagene, La Jolla, CA), with the template being a small portion of the Ca_V2.1 α_1A cDNA, which spanned the IQ domain. The mutant IQ domain was then cloned into a pristine α_1A subunit in pcDNA3, via unique upstream BstE II and downstream Xba I restrictions sites, which flanked the IQ domain. For the chimeric channels in Figure 4, the caboxy tail of the Ca_V2.3 α_1E subunit was replaced by the N-terminal third of the carboxy tail (containing EF-hand and IQ regions) derived from various combinations of Ca_V1.2 α_1C and Ca_V2.3 _E subunits. Thus, amino acids 1–1712 of the Ca_V2.3 α_1E subunit served as a ‘backbone’ for all chimeric channels. For construct 1 (Figure 4C), PCR was used to amplify cDNA encoding amino acids 1513–1670 of the Ca_V1.2 α_1C subunit, followed by a stop codon. This product was cloned into unique upstream Xho I and downstream Xba I sites of the Ca_V2.3 α_1E subunit. The upstream Xho I site in the PCR product was introduced as a silent mutation at the upstream end of the PCR product. For construct 2 (Figure 4C), amino acids 1818–1982 of the Ca_V2.1α_1A subunit was introduced into the Ca_V2.3 α_1E subunit by the same strategy and restriction sites. For constructs 3–6 (Figure 4C), overlap extension PCR (Ho et al., 1989) was used to generate the various carboxy tail inserts, which were introduced into the Ca_V2.3 α_1E subunit by the same strategy and restriction sites. The detailed amino acid composition of constructs 1–6 are presented in Supplemental Data 2. Throughout, all portions of constructs subject to PCR or QuickChange were confirmed in their entirety by sequencing.

Energy calculations for point mutations

The web-based computational algorithm Robetta (http://www.robetta.org/alascansubmit.jsp) was used to predict the change in ligand-receptor (i.e., IQ domain and Ca²⁺/CaM) binding energy produced by point mutation of interface residues (Kortemme et al., 2004). These algorithms accommodate changes not only to alanine, but also to other amino acids (e.g., Figure 5, M[7minus;1]T and Y[+4]E). For mutations to alanine, the default operational mode of the algorithm was used. For mutations from alanine to non-alanine residues (A→X), we modified the target residue in the submitted structure file, so as to rename the target residue as ‘X’ and delete all side-chain atoms. This change triggered the computational algorithm to build in the missing atoms for the desired mutant residue (X), and then optimize rotomer conformation. The calculated ΔΔG(X→A) then corresponded to an X→A mutation. The energy for the desired A→X mutation is then determined as ΔΔG(A→X) = −ΔΔ G(X→A). For Y→X mutations, where neither X nor Y is alanine, a concatenation of the former manipulations was used. Specifically, ΔΔG (Y→X) = ΔΔG(Y→A) + ΔΔG(A→X). Calculations were generated based on our structure of Ca²⁺/CaM in complex with the IQ domain of Ca_V2.3, rather than for the Ca_V2.1 analog, owing to the advantages of the higher resolution in the former structure (2.1 Å versus 2.6 Å).