|Home | About | Journals | Submit | Contact Us | Français|
Voltage-gated calcium channels (CaVs) are large, multisubunit complexes that control cellular calcium entry. CaV pore-forming (CaVα1) and cytoplasmic (CaVβ) subunits associate through a high-affinity interaction between the CaVα1 α-interaction domain (AID) and CaVβ α-binding pocket (ABP). Here, we analyze AID-ABP interaction thermodynamics using isothermal titration calorimetry (ITC). We find that commensurate with their strong sequence similarity, all CaV1 and CaV2 AID peptides bind CaVβ with similar nanomolar affinities. Although the AID-ABP interface encompasses twenty-four sidechains, alanine-scanning mutagenesis reveals that the binding energy is focused in two complementary hotspots comprising four deeply-conserved residues. Electrophysiological experiments show that hotspot interaction disruption prevents trafficking and functional modulation of CaV1.2 by CaVβ. Together, the data support the primacy of the AID-ABP interface for CaVα1-CaVβ association, underscore the idea that hotspots dominate protein-protein interaction affinities, and uncover a target for strategies to control cellular excitability by blocking CaVα1-CaVβ complex formation.
Voltage-gated calcium channels (CaVs) are the molecules that define excitable cells (Hille, 2001). These channels play critical roles in cardiac action potential propagation, neurotransmitter release, muscle contraction, calcium-dependent gene-transcription, and synaptic transmission (Catterall, 2000). High-voltage activated CaVs (CaV1.x and CaV2.x) are multiprotein complexes that are composed of four principal components (Van Petegem and Minor, 2006): a transmembrane, pore-forming CaV α1-subunit (CaV α1) (Catterall, 2000), a cytoplasmic CaVβ-subunit (CaVβ) (Dolphin, 2003a), a single pass transmembrane subunit, CaVα2δ (Cantí et al., 2003), and calmodulin (CaM) (Pitt, 2007).
The interaction between the CaVα1 and CaVβ subunits shapes channel gating properties (Colecraft et al., 2002; Hullin et al., 2003), influences channel trafficking (Beguin et al., 2001; Bichet et al., 2000), affects regulation by neurotransmitter receptors through G-protein βγ subunit activation (Dolphin, 2003b), and tunes channel sensitivity to drugs (Hering, 2002). The association of CaVα1 and CaVβ subunits is critical for proper channel maturation and cell surface expression (Bichet et al., 2000; Takahashi et al., 2005) and thereby can profoundly influence cellular excitability properties. The physiological importance of the CaVα1-CaVβ interaction and its influence on channel function is highlighted by in vivo studies of CaVβ mutants and knockout mice in which loss of this interaction leads to reduced voltage-gated calcium channel currents (Gregg et al., 1996; Murakami et al., 2002; Murakami et al., 2003; Namkung et al., 1998; Weissgerber et al., 2006) and a host of physiological defects (Burgess et al., 1997; Gregg et al., 1996; Murakami et al., 2002; Murakami et al., 2003; Weissgerber et al., 2006). Considering its central role in CaV function, a detailed knowledge of the CaVα1-CaVβ interaction is critical for understanding how CaVβs control and shape calcium channel action.
The principal CaVα1-CaVβ interaction site on the pore-forming subunit is a conserved eighteen residue sequence in the I–II loop called the α-interaction domain, ‘AID’ (Pragnell et al., 1994; Witcher et al., 1995). High-resolution X-ray crystallographic analysis of CaVα1 AID-CaVβ complexes (Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004) defined the molecular architecture of the AID-CaVβ interaction. CaVβs are built from two interacting conserved domains, an SH3 and nucleotide kinase (NK) domain that make the functional core (Opatowsky et al., 2003) that is flanked by variable regions. The structural work showed that the AID forms an α-helix that binds a deep CaVβ pocket, termed the ‘α-binding pocket’ (ABP) (Van Petegem et al., 2004). Previous assessment of the AID-CaVβ interaction affinity by a variety of methods suggests that binding is strong (Kd~nM) (Bell et al., 2001; Butcher et al., 2006; Canti et al., 2001; Geib et al., 2002; Opatowsky et al., 2003) and coupled to AID α-helix formation (Opatowsky et al., 2004).
A number of outstanding questions remain regarding the AID-CaVβ interface. First, AID sequences from different CaV1 and CaV2 isoforms show some variation. While the measured affinities reported for a few individual AID-CaVβ pairs are similar, there are no studies reporting a systematic evaluation of all CaV1 and CaV2 AID isotypes. Therefore, it is unclear whether the isoform specific AID sequence variation affects the AID-CaVβ binding affinity in a way that could influence channel function. Second, despite the apparent key role for the AID-CaVβ interaction in proper CaV function (Dolphin, 2003a), studies of a CaV2.1 AID deletion mutant have challenged the idea that the AID-CaVβ interface is essential for channel modulation (Maltez et al., 2005). Also in seeming opposition to the high affinity of the AID-CaVβ interaction reported in prior biochemical studies (Bell et al., 2001; Butcher et al., 2006; Canti et al., 2001; Geib et al., 2002; Opatowsky et al., 2003), some functional studies have suggested that the AID-CaVβ interaction may be reversible under certain situations (Hidalgo et al., 2006; Restituito et al., 2001). Finally, although no molecules that block the AID-CaVβ interaction have yet been validated biochemically, discovery of small molecule modulators of the AID-CaVβ interaction may provide a new way to control cellular excitability and open a path towards the development of novel CaV antagonists (Young et al., 1998).
There is much interest in the design of molecules that disrupt specific protein-protein interactions. The challenge is that most protein-protein interaction surfaces are large (750–1500 Å2) and are amenable to small molecule interference only if a subset of the interaction surface is necessary for binding affinity (Arkin and Wells, 2004). The AID-ABP interaction buries ~730 Å2 of ABP surface area (Van Petegem et al., 2004) and is composed of amino acids that are highly conserved. Thus, sequence comparisons provide scant information regarding how essential any of the sidechains might be for forming the AID-ABP interaction. It is not possible from the structural data alone to determine which portions of the AID-ABP protein-protein interface are essential contributors to the high-affinity interaction and how much any particular interaction actually contributes to binding affinity. Given the conflicting views of the importance of the AID-CaVβ interaction for CaV function and the potential for using block of this interaction for the development of new CaV antagonists, we set out to define the energetic and structural underpinnings of the AID-CaVβ interaction. Our data reveal that the AID-ABP binding energy is focused into two complementary hotspots comprising four amino acids and that both hotspots are essential for CaVβ modulation of channel function.
CaV1 and CaV2 α1-subunits have similar AID regions through which they interact with CaVβs (Table 1). We used ITC to characterize the interaction of the AIDs from each of the CaV1 and CaV2 isoforms with CaVβ2a. This approach allows the direct measurement of the thermodynamic parameters of the binding reaction and gives an unmatched level of accuracy and resolution (Velazquez Campoy and Freire, 2005). We used purified AID peptides made biosynthetically and a CaVβ2a construct that contained the functional core regions identified in the high-resolution crystal structure (Van Petegem et al., 2004). ITC experiments comparing AID binding to the crystallized CaVβ2a construct, composed of two non-covalently associated domains (Van Petegem et al., 2004), and a unimolecular construct bearing a glycine linker between the SH3 and NK domains gave identical affinities (Table 2). Thus, we used the unimolecular CaVβ2a construct for experiments with different AID isoform peptides and as the basis for our CaVβ2a alanine scan (described below) as it was easier to purify.
Measurement of the affinities of the CaV1 and CaV2 AID isoforms for CaVβ2a by ITC shows that all AID isoforms bind in the low nanomolar range (Figure 1A–C, Table 2). In all cases the interaction is driven by a large, enthalpic component. The similar thermodynamic properties displayed by all of the AID binding reactions are consistent with the high degree of conservation among the AID residues that make contacts with the ABP (Van Petegem et al., 2004). While still having a tight binding affinity (Kd = 53.5 nM), the AID peptide for CaV2.3 bound less avidly than the other AID isoforms. This difference appears to originate from a non-conservative substitution at one of the positions that contacts the ABP, (corresponding to CaV1.2 L483, see below). Exchange of CaV2.3 R for M, as found at the equivalent CaV2.1 position (Table 1), increased the binding affinity ~6 fold (Kd = 8.6 nM) (Table 2) and brought the CaV2.3 AID affinity on par with the other isoforms.
The affinities measured for CaV1.2 and CaV2.1 AIDs by ITC correspond well to those measured by other biophysical methods: CaV1.2 AID-CaVβ2a, (Kd = 16.1 nM, measured by fluorescence polarization) (Opatowsky et al., 2003), and GST-CaV2.1 AID-CaVβ4 (Kd = 9.3 nM, measured by surface plasmon resonance) (Geib et al., 2002). The ITC values are also in close agreement with measurements reported using CaVα1 I–II loops, which bear the AID at the N-terminal portion of the loop: CaV1.3 I–II linker-CaVβ1b (Kd = 10 nM) (Bell et al., 2001), CaV2.2 I–II linker-CaVβ1b (Kd = 21nM, and 13.7 nM) (Bell et al., 2001; Butcher et al., 2006), CaV2.2 I–II linker-CaVβ3 (Kd = 26 nM) (Canti et al., 2001). The similarity between the Kd values for the AID alone measured by calorimetry and the Kd values for the I–II loops as measured by surface plasmon resonance provides strong support for the idea that the AID is the only high affinity binding site on the I–II loop and that other proposed interactions (Maltez et al., 2005) contribute marginally, if at all, to the CaVα1 I–II linker-CaVβ binding affinity.
Having established that only minor differences exist among the binding energetics for different high-voltage CaV AID isoforms to CaVβ, we sought to define which AID-ABP interface residues were essential for the high-affinity interaction by employing alanine-scanning mutagenesis of the CaV1.2 AID and CaVβ2a ABP together with ITC measurements. Alanine substitution deletes all interactions made by sidechain atoms beyond the β-carbon and provides a means to test the contribution of the mutated sidechain to the binding energetics of complex formation (Wells, 1991).
The AID forms an α-helix that binds to the ABP, which is located in the CaVβ NK domain (Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004), through interactions made by ten AID sidechains (Van Petegem et al., 2004). We mutated each of the CaV1.2 AID contact residues to alanine and measured the effect on the AID-CaVβ2a interaction by ITC (Figure 1D, E, G–I, Table 3). Mutation of residues at the N-terminal (Q429A, L430A, and D433A) and the C-terminal (Q443A) ends of the AID helix to alanine did not affect binding affinity (Table 3). In contrast, alanine mutation of each of the six ABP-contacting residues in the middle of the AID α-helix (L434A, G436A, Y437A, L438A, W440A, and I441A) perturbed binding substantially. We observed the greatest effects (|ΔΔG| > 3.5 kcal mol−1) at three CaV1.2 AID positions, Y437A, W440A, and I441A. W440A caused the most severe affinity decrease, reducing the affinity to a level that was beyond the dynamic range of our measurements (|ΔΔG| > 4.3 kcal mol−1)(Figure 2G).
In addition to the loss of interactions in the complex, some of the alanine substitution effects on binding could result from changes in the helical character of the mutant AID peptides. We used circular dichroism to examine whether the AID peptide has any substantial helical character under the conditions of the ITC experiment. In agreement with previously reported studies (Opatowsky et al., 2004), we find that the isolated AID peptide is essentially unstructured (Supplementary Figure 1). While binding and AID helix formation are linked (Opatowsky et al., 2004), in the context of a binding reaction that is favored by >10 kcal mol−1, small changes in the helical content of the largely disordered AID peptide are unlikely to have any significant effect. Thus, one can confidently attribute the changes in binding affinity to loss of AID sidechain interactions from the complex.
The AID alanine scan data reveal a striking pattern when considered in the context of the AID-CaVβ structure (Figure 2). The residues that carry the greatest share of the binding energy (|ΔΔG| > 3.5 kcal mol−1), Y437, W440, and I441, form a small, well-defined hotspot (Figure 2A) that is buried deep in the ABP and flanked by residues of lesser energetic importance (Figure 2B & C). The large thermodynamic contribution of Y437, W440, and I441 for AID-CaVβ binding affinity agrees with prior work that identified these three AID positions as key binding determinants for the AID-CaVβ interaction (De Waard et al., 1996). The concentration of AID binding energy into a small, contiguous portion of the AID three dimensional structure is a striking manifestation of how a limited set of residues can dominate the binding affinity of a large protein-protein interface (Bogan and Thorn, 1998; Clackson and Wells, 1995).
Removal of the conserved AID tyrosine sidechain hydroxyl by Y→F mutation has been reported to reduce the affinity of the AID-CaVβ interaction (De Waard et al., 1996; Witcher et al., 1995). Because we observed such a large binding affinity decrement for Y437A (ΔΔG = − 3.98 kcal mol−1), we were interested in quantifying the effect of Y437F on the AID-ABP interaction. Y437F decreases the AID-CaVβ interaction strength by a remarkable amount, ~70 fold, affects both enthalpic and entropic binding terms, and makes the entropic term unfavorable, an effect that few other mutants have (Figure 1F, Table 3). The Y437 sidechain hydroxyl is coordinated by two water molecules and is part of a water-mediated hydrogen-bonding network that is largely built from backbone hydrogen bonding groups. The waters that bind the Y437 hydroxyl are present in two, independently determined high-resolution CaV1.2 AID-CaVβ2a complex crystal structures (Opatowsky et al., 2004; Van Petegem et al., 2004) (Figure 2D, Supplementary Table 1) and absent from both high-resolution apo-CaVβ2a structures (Opatowsky et al., 2004; Van Petegem et al., 2004). The large energetic penalty, 60.8% of the total binding energy contributed by the entire aromatic sidechain, for the simple removal of the tyrosine hydroxyl by the Y→F mutation appears to arise from the loss of the central moiety in the entire hydrogen bond network. Thus, formation of the hydrogen bond network surrounding the Y437 hydroxyl plays a crucial role in the AID-CaVβ interaction. The importance of the hydrogen bonding network surrounding this AID tyrosine is further reinforced by the strict conservation of tyrosine at this AID position among organisms that are separated by many hundreds of million years of evolution (Table 4).
The sidechains lining the CaVβ ABP are so strongly conserved (Table 4) that interspecies comparison offers few clues into the relative importance of any individual residue for AID binding. In order to understand which portions of the ABP are critical for the AID-ABP interaction, we also performed an alanine scan of the fourteen CaVβ residues that make sidechain-mediated interactions with the AID (Figure 3). Sidechain truncation to alanine of nine of the ABP residues had either no effect, |ΔΔG| ≤ 0.5 kcal mol−1 (S345A, R351A, N390A, Q391A, and E393A), or modest effects, 0.5 < |ΔΔG| ≤ 1.5 (M244A, V341A, V348A, S355A), on AID binding affinity (Table 3). Five alanine mutants perturbed the binding affinity by greater than tenfold, |ΔΔG| > 1.5 kcal mol−1 (M245A, I343A, L352A, R356A, L392A), with M245A having the largest impact, 250-fold, ΔΔG = −3.16 kcal mol−1. The Kd values for M245A and L392A measured by ITC (1.148 μM and 107 nM, respectively) are consistent with those previously reported by fluorescence polarization assays (1.7 μM and 340 nM, respectively) (Opatowsky et al., 2004). To test whether any of the mutants that caused substantial effects on binding perturbed the binding constant by virtue of disrupting the native CaVβ2a structure, we compared the far UV circular dichroism (CD) spectra of M245A, I343A, L352A, R356A, and L392A with the CD spectrum of the wild-type protein (Figure 4A). All five CaVβ2a mutants that affect AID binding affinity have CD spectra that are similar to wild-type in shape and magnitude. These data support the idea that the alanine mutations affect binding through deletion of important sidechain contacts in the AID-ABP interface and not through disruption of the overall CaVβ structure.
Examination of the CaVβ alanine scan perturbations reveals a conspicuous pattern. The ABP surface contains a single binding hotspot for the AID that has M245 at the center (Figure 4B). The ABP hotspot provides a remarkably complementary surface for interaction with the three-dimensional structure formed by the AID hotspot residues (Figure 4C). M245 forms a ridge in the ABP that engages the concave surface formed by AID hotspot residues Y437, W440, and I441 that precisely matches the sidechain conformation of CaVβ M245. This ridge is preformed; M245 adopts the identical conformation in the apo-CaVβ2a ABP (Van Petegem et al., 2004).
Given the questions regarding the importance of the AID-CaVβ interaction (Maltez et al., 2005) and prior demonstration that the functional defects caused by single AID mutants can be overcome through increased CaVβ expression (Butcher et al., 2006), we decided to revisit the functional importance of the AID-ABP interaction in light of our hotspot identification. Based on our ITC data, we made a triple mutant in the CaV1.2 AID (Y437A/W440A/I441A, denoted ‘HotA’). Assuming strict additivity for the alanine mutants, the Kd of AID binding for this mutant should be greater than 6 M. Thus, the HotA mutant combination should effectively eliminate the AID-CaVβ interaction.
Expression in Xenopus oocytes of ‘HotA’ produced barium currents that had very similar properties to wild-type CaV1.2 channels expressed in the absence of CaVβ2a (Figure 5A–C). As reported previously (Neely et al., 1993; Perez-Reyes et al., 1992; Yamaguchi et al., 2000), we observed that co-expression of CaVβ2a with wild-type CaV1.2 in oocytes increases current amplitude and is accompanied by a ~25 mV hyperpolarization of the activation V1/2 (Figure 5A–C). In contrast, co-expression of HotA with CaVβ2a failed to increase the size of channel currents when equivalent amounts of RNA were compared (Figure 5A, B). In some batches of oocytes (two of four), we observed a slight (~5–10 mV) hyperpolarization of the activation V1/2 when HotA was co-injected with large amounts of CaVβ2a RNA. These shifts did not appear to be caused by simple co-expression of a soluble protein as co-expression with the unrelated channel modulator, KChIP1, failed to induce V1/2 changes (data not shown). These results may indicate that some AID-independent modulation by CaVβ2a is possible; however, the V1/2 changes are small, occur in the absence of the stimulation of current amplitude, and are not a robust effect making it difficult to attribute them to a direct effect of CaVβ. Their significance and origins remain to be elucidated. The gross insensitivity of the HotA mutant to CaVβ2a modulation clearly demonstrates that the AID-CaVβ2a interaction is necessary for increasing current amplitudes and complete hyperpolarization of the I–V curve.
To examine the role of the affinity of the AID-CaVβ2a interaction in channel modulation, we characterized the functional properties of the CaVβ hotspot mutation M245A, which greatly decreases but does not eliminate the AID-CaVβ2a interaction. In agreement with previous findings (Maltez et al, 2005), co-expression of CaV1.2 and CaVβ2a-M245A produced channels with current amplitudes and biophysical parameters that were equivalent to co-expression of CaV1.2 with the same amount of wild-type CaVβ2a cRNA (data not shown).
As there is a substantial impact of M245A on AID binding (250-fold reduction, Table 3), we sought an explanation for why a mutation that causes such a gross defect in binding affinity lacked a clear functional effect. CaVβ2a is the only CaVβ that bears an N-terminal palmitoylation site that helps anchor it to the plasma membrane (Chien et al., 1996). Because the lipid anchor is likely to increase the local concentration of CaVβ2a at the membrane, we reasoned that it might be responsible for our inability to observe functional effects in the CaVβ2a M245A mutant. To test this hypothesis, we made a CaVβ2a mutant that lacks the requisite cysteines at the palmitoylation site (C3S/C4S, denoted ‘ssCaVβ2a’) and that has been shown to prevent CaVβ2a lipid modification (Chien et al., 1996). ssCaVβ2a was capable of stimulating current amplitude and hyperpolarizing the I–V relationship but required the injection of larger amount of cRNA compared to CaVβ2a in order to surmount the loss of the membrane anchors (Figure 5D and E). Examination of the M245A mutation in the ssCaVβ2a background unveiled the functional impact caused by the reduction of AID binding affinity (Figure 5D, E). At concentrations where ssCaVβ2a could modify channel properties, M245A failed to stimulate current amplitude and caused only minor modification to the biophysical parameters. Together, the data indicate that the defect in binding energy caused by the M245A mutation can be overcome by increasing the effective concentration (Jencks, 1981) of the protein at the membrane, either by propinquity effects resulting from the membrane anchor or from increased expression levels, and highlight the role of the high affinity AID-CaVβ interaction for CaVβ subunits lacking additional membrane anchors. This simple explanation had been previously overlooked (Maltez et al., 2005). Given the impotence of CaVβ in modulating the trafficking and biophysical properties the pore forming subunit in the context of either HotA or M245A, our data strongly suggest that any other interactions between CaVβ2a and the pore forming subunit are quite weak. Contrary to alternative proposals (Maltez et al., 2005), the data indicate that the AID-CaVβ2a interaction is essential for the two most prominent modes of CaVβ modulation of the pore forming subunit, the hyperpolarizing the I–V relationship by CaVβ subunits and the stimulation of current amplitude.
Association of CaVα1 pore-forming subunits from CaV1 and CaV2 isoforms and CaVβ cytoplasmic subunits is essential for proper trafficking and functional fidelity of high-voltage activated calcium channels (Dolphin, 2003a). Structural studies (Chen et al., 2004; Opatowsky et al., 2004; Van Petegem et al., 2004) show that the conserved eighteen-residue sequence in the cytoplasmic I–II loop of the pore forming subunit called the α-interaction domain, ‘AID’, binds to a deep hydrophobic cleft on the CaVβ NK domain, named the ABP (for α-binding pocket) (Van Petegem et al., 2004). The AID-ABP interaction buries ~730 Å2 of ABP surface area and comprises interactions among twenty-four sidechains that are similar regardless of the CaVα1 or CaVβ isoforms involved. Accordingly, despite some divergence of AID sequences among CaV1 and CaV2 subunits (Table 1), our measurements show that the binding affinities do not vary greatly (Table 2).
Understanding the nature of protein-protein interfaces is an important general problem (Reichmann et al., 2007). Even though the AID-CaVβ structures define the atomic contacts that form the AID-ABP protein-protein interface, it is not possible using structural data alone to determine which portions of the AID-ABP interface are essential contributors to the high-affinity interaction or how much particular interactions might contribute to binding affinity. Previous work suggests that the majority of protein-protein binding affinity is often contributed by hotspot residues that constitute a limited set of the total interactions observed in an interface (Bogan and Thorn, 1998; Clackson and Wells, 1995; Desrosiers and Peng, 2005). The definition of such energetic hotspots has important implications for understanding protein-protein interactions and for the development of small molecules that might be able to target such hotspots specifically and prevent association (Arkin and Wells, 2004). Our studies identify a pair of complementary hotspots that restrict the energetically important interactions in the relatively large AID-ABP binding interface to those made by a mere four crucial amino acids, CaV1.2 Y437, W440, I441, and CaVβ2a M245 (Figures 2D and and4C).4C). These hotspot residues are essentially invariant across hundreds of millions of years of metazoan evolution from humans to cnidarians, the oldest group of organisms to possess a nervous system (Table 4). This remarkable degree of conservation suggests that there is intense selective pressure to maintain the high affinity nature of the CaVα1-CaVβ interaction in diverse settings.
Initial studies made it clear that the AID Tyr-Trp-Ile triplet had an important role in mediating the AID-CaVβ interaction (De Waard et al., 1996). Further examination of AID interaction residues in a variety of CaVα1 isoforms has provided strong evidence for a direct link between the ability of CaVβ to bind the AID and its ability to modulate channel biophysical properties. In accord with the results reported here, data from Neely and colleagues support the importance of W440 as the W440S mutation abolishes both binding and modulation of CaV1.2 by CaVβ2a (Hidalgo et al., 2006). In agreement with the distribution of AID-ABP binding energy that we observe, data from Parent and colleagues have shown that alanine substitution of the CaV2.3 AID at positions equivalent to CaV1.2 Q429, L430, D433, L434, and G436 do not affect binding or modulation by CaVβ3 (Berrou et al., 2005), whereas CaV2.3 mutants at positions equivalent to CaV1.2 Y437, W440, I441 disrupt both binding and modulation (Berrou et al., 2001; Berrou et al., 2005).
Studies of CaV2.2 AID mutants by Dolphin and colleagues also support the importance of the Y437 and W440 positions for binding affinity and channel modulation (Butcher et al., 2006; Leroy et al., 2005). Mutation of CaV2.2 AID residues equivalent to CaV1.2 Y437 and W440 to Ser and Ala, respectively, reduced binding and the ability of CaVβ1b to modulate the channel. Importantly, the modulatory deficits of the Tyr→Ser mutation could be overcome by titrating the CaVβ1b expression level, demonstrating that some binding affinity deficiencies can be overcome by increased subunit concentration (Butcher et al., 2006), an idea that is also supported by our experiments with the CaVβ mutant M245A. The CaV2.2 AID mutation equivalent to W440A abolished binding of CaVβ isoforms to both the AID and I–II loop and eliminated the cell surface expression effects and biophysical modulation by isoforms lacking N-terminal palmitoylation, CaVβ1b and CaVβ3 (Leroy et al., 2005). Co-expression of palmitoylated CaVβ2a with CaV2.2 W440A caused some stimulation of surface expression, clear shifts in the inactivation V1/2, and small shifts (~−4mV) in the activation V1/2. No such shifts were present with the non-palmitoylated CaVβ2a mutant, suggesting that palmitoylation can overcome some of the functional effects caused by binding affinity reduction. In contrast to these studies of single AID point mutants, we find that the biophysical properties and current amplitudes of the HotA mutant, which carries alanine substitutions at Y437, W440, and I441, is insensitive to increased CaVβ2a expression. CaV1.2-CaVβ2a co-expression produces a much larger hyperpolarizing shift in the activation V1/2, ~−25mV (Figure 5C), than the shift elicited by CaV2.2-CaVβ2a co-expression (~4mV) (Leroy et al., 2005) when compared to channels expressed in the absence of CaVβ2a. The discrepancy with the CaV2.2 channel studies may be due to a small residual affinity between the W440A AID and CaVβ2a that can be compensated by increasing the CaVβ2a local concentration or because the CaVβ2a-induced shifts of CaV2.2 are not large enough to provide a robust metric of the functional influence of CaVβ2a.
A number of studies have generated some disagreement about the relative importance of the AID for modulation of channel function. Pitt and colleagues have suggested that many of the properties conferred on the pore-forming subunit by CaVβ are independent of the AID (Maltez et al., 2005). CaV2.1 channels lacking the AID show a modest modulation of inactivation rates (~1.7 fold) in the presence of CaVβ2a (Maltez et al., 2005), although these channels do not display the normal current enhancement that is characteristic of CaVα1-CaVβ2a interactions. Co-expression of CaVβ2a with CaV2.1 channels lacking an AID segment leads to a V1/2 for activation that is similar to that reported for CaVβ2a co-expressed with wild-type CaV2.1 (Maltez et al 2005). Our results with the CaV1.2 HotA mutant stands in marked contrast to this prior work (Figure 5C). The CaV2.1 AID deletion removes key residues that are known to be important for modulation of inactivation (Dafi et al., 2004; Herlitze et al., 1997) and may further produce a defect in the channel structure that unintentionally affects channel properties. Either or both of these issues may be the source of the apparent discrepancy with our results. Our alanine substitutions, which merely delete the sidechains of residues essential for the AID-CaVβ interaction but do not rearrange the amino acids that are proximal to the IS6 segment, are less likely to cause major changes to the local structure. Additionally, the excellent agreement between our ITC measurements for the AID and prior studies on I–II loop binding (Bell et al., 2001; Butcher et al., 2006; Canti et al., 2001) provides strong support for the idea that the AID is the only high affinity binding site on the I–II loop and that other proposed interactions (Maltez et al., 2005) contribute marginally, if at all, to the binding affinity.
Recently, it has been proposed that small fragments of CaVβ2 arising from putative splice variants are capable of interacting with the α1-subunit and can modulate channel activity and membrane targeting in the absence of the functional conserved CaVβ core (Harry et al., 2004). Because the AID hotspot is strictly necessary for increased channel expression and channel activation, the proposed effects and relevance of the non-conserved CaVβ regions in the absence of a functional core (Harry et al., 2004) remain questionable. Claims that the AID-CaVβ in vivo affinity is orders of magnitude weaker than the in vitro interaction (Hidalgo et al., 2006) also seem difficult to reconcile with the overwhelming biophysical data that demonstrate the high affinity and critical nature of the interaction. Taken together, the studies of mutants at conserved AID anchor positions in CaV1.2, CaV2.2, and CaV2.3 provide strong evidence that the conserved AID-CaVβ interaction at the ABP is critical for both cell biological and biophysical modulation of CaVs by CaVβ subunits.
The observation that different CaVβs, which all share a conserved core containing the SH3 and NK domains, cause different biophysical effects on CaVα1 subunits (Dolphin, 2003a) requires some mechanism besides the conserved AID-ABP interaction through which CaVβs influence channel conformational changes. Consideration of the likely three-dimensional context of CaVβ with respect to CaVα1 necessitates some other interaction points between CaVβ and the pore-forming subunit besides the AID-ABP high affinity site. Yang and coworkers have shown that the NK domain alone can stimulate current amplitudes and have partial effects on channel activation and inactivation properties (He et al., 2007). The full modulatory effects, however, required the variable N-terminal domain that precedes the SH3 domain, and the variable HOOK domain that bridges the SH3 and NK domains. These results provide evidence that other CaVβ domains are important for influencing channel conformational changes associated with channel activity. Sites of lower affinity CaVα1-CaVβ interactions have been proposed for the CaVα1 C-terminal domain (Walker et al., 1998) and within the I–II loop (Maltez et al., 2005). The nature of such interactions remains unresolved. Crystallographic studies have suggested one plausible site on the SH3 domain (Van Petegem et al., 2004; Van Petegem and Minor, 2006) and functional studies point to the HOOK domain as another candidate (He et al., 2007). Nevertheless, biophysical analysis of the CaVα1-CaVβ interaction has failed to find any real difference in the affinity of the AID and the I–II loop. It seems likely that there are three-dimensional requirements for such interactions that depend on the context of the entire channel. Certainly, any CaVα1-CaVβ interactions that have a modest intrinsic affinity (e.g. micromolar to millimolar) could gain potency due to the high affinity anchoring of CaVβ to CaVα1 through the AID-ABP interaction. The present challenge is to develop ways to map the intrinsically low-affinity interactions between other portions of CaVβ and CaVα1 that shape the diverse modulatory effects of the different CaVβ isoforms but that may be too weak to discover by conventional biochemical affinity assays.
Because of the central importance of CaVs to the excitable properties of cells in the nervous system and heart, there is a strong interest in the development of novel calcium channel modulators. Previous work suggested that it may be possible to modulate CaV activity by using a small molecule to interfere with the CaVα1-CaVβ interaction, although whether the identified benzyl-pyrimidine ether (Young et al., 1998) truly blocks the CaVα1-CaVβ interaction is unknown. Our identification of a well-defined hotspot for the CaVα1-CaVβ interaction in which most of the interaction energy is focused into a small portion of the AID-ABP interface provides an attractive target for the development of novel small molecules that modulate cellular excitability by interfering with CaV biogenesis.
The AID segments of CaV1 and CaV2 family members were cloned into either the pSV272 or pet28HMT vector (Van Petegem et al., 2004), containing, in sequence, a hexahistidine tag, maltose binding protein, a cleavage site specific for the tobacco etch virus protease (TEV), and the AID segment. The AID segments consist of 18 amino acids each (Table 1), originating from clones for rabbit CaV1.1, human CaV1.2, rabbit CaV2.1, rabbit CaV2.2 and rat CaV2.3. A construct of rat CaVβ2a, consisting of the SH3 domain (residues 17–138) and the NK domain (residues 203–425) separated by a serine residue, was cloned into the pET28HMT vector. All mutants were generated using the quikchange protocol (Stratagene).
All proteins were expressed in Escherichia coli BL21(DE3)pLysS grown in 2xYT media at 37 °C. Cells were lysed in a buffer containing 10% sucrose, 150 mM KCl, 10 mM β-mercaptoethanol, 1 mM EDTA, 1 mM PMSF, 10 mM Tris, pH 8.8.
For the AID constructs, the HMT fusion protein was first loaded onto a Poros20MC affinity resin (Perseptive Biosystems) in 250 mM KCl, 10mM K+-phosphate pH 7.3 (buffer A) and eluted in successive 30% and 100% steps of buffer B (buffer A + 500 mM imidazole pH 7.3). The eluate was dialyzed against a low ionic strength buffer containing 10 mM KCl, 20 mM Tris, pH 8.8 (buffer C), loaded onto a Hiload Q column (Pharmacia), and eluted with 1 M KCl, 20 mM Tris 8.8 (buffer D) using a linear gradient of 11–50% buffer D over 12 column volumes. The protein fractions were concentrated to ~2ml (Amicon, 10K cutoff) and incubated overnight with 50 μl of 30.5 μM TEV protease at room temperature. The cleaved sample was run on a TSK2000 gel filtration column (Tosoh Biosep) in a buffer of 250 mM KCl, 10 mM potassium phosphate, pH 7.3. Fractions containing the AID peptide were purified further by HPLC using a C18 HPLC column (Vydac) and a 0.1% min−1 water/acetonitrile linear gradient in 0.1% (v/v) trifluoroacetic acid. Following HPLC samples were lyophilized and stored until use. Sample identity was confirmed using MALDI-TOF MS.
For the CaVβ2a constructs, protein was loaded onto Poros20MC resin in buffer A, and eluted in buffer B. The sample was dialyzed against buffer A, and incubated with TEV protease overnight. Following cleavage, the material was applied to a tandem series of columns (Amylose/POROS) to remove the His-tagged TEV protease, cleaved HMT tag, and any uncleaved material. The protein of interest was collected in a wash step with 6% buffer B. In most cases the material was pure. For some of the preprations, the cleaved material required one more purification step to remove minor impurities. In these cases the protein was dialyzed (1K cutoff tubing) against 10mM KCl, 20mM MES pH6.3 (buffer E), and run on a Hiload SP column (Pharmacia) and eluted with a gradient of 30 to 70 % buffer F (1M KCl; 20mM MES pH 6.3) over 12 CV.
Samples were concentrated and dialyzed against 150 mM KCl, 10 mM potassium phosphate, pH 7.3. Samples were degassed for 5 min and titrations were performed on a VP-ITC calorimeter (MicroCal) at 15 °C. For all measurements, injections of 7–10 μl of AID peptide were titrated into 1.4 ml of CaVβ2a to a stoichiometric ratio of 2–2.5. Data were processed with MicroCal Origin 7.0.
Protein concentrations were determined by absorbance (Edelhoch, 1967). Each ITC experiment was repeated with different batches of purified protein and yielded similar thermodynamic parameters and stoichiometry values. For experiments with lower affinities we also titrated AID into buffer to adjust the baseline. For high affinity measurements (nM affinities), the final titration points were used to estimate the baseline.
CD spectra were measured for all CaVβ2a mutants using an Aviv Model 215 spectropolarimeter (Aviv Biomedical) equipped with a peltier device. Wavelength scans from 320 nm to 185 nm were taken at 15° C in a 2 mm path length quartz cuvette. Molar ellipticity per residue of the buffer-subtracted CD spectrum using the following relationship: [θ] = mdeg *100/(M*l*NR), where [θ] is the molar ellipticity per residue in deg-cm2 (dmol-res)−1, mdeg is the experimental ellipticity in millidegrees, M is the molecular mass of the protein, c is the protein concentration in millimolar, l is the cuvette path length in cm, and NR is the number of residues in the protein (349 for the linked CaVβ2a construct). All CaVβ spectra were measured at a protein concentration of 10 μM. Protein concentrations were determined by absorbance (Edelhoch, 1967).
Constructs for electrophysiology consisted of human CaV1.2 (splice variant α1c77) in pcDNA3.1(+)/hygro (Invitrogen) and CaVβ2a in pGEM (Promega). Mutants of the α1c77 and β2a subunits were made using the QuikChange protocol (Stratagene). RNA transcripts were prepared using a T7 mMessage mMachine kit (Ambion). 50 nl of a complementary RNA mixture containing 50 nM CaV1.2 α1c77 and 1–25nM CaVβ2a was microinjected into Xenopus oocytes, which were then kept at 18 °C in ND96 medium supplemented with penicillin and streptomycin. Recordings were performed 2–5 days after injection. In order to increase the reproducibility, recordings comparing the HotA with wild-type CaV1.2 were performed on the same batch of oocytes. Recordings comparing the CaVβ2a, ssCaVβ2a, CaVβ2a M245A, and ssCaVβ2a M245A were performed on a separate batch of oocytes, with all recordings performed 4 days after injection. Before recording, oocytes were injected with 25–50 nl 100 mM BAPTA to minimize contaminating calcium-activated chloride current. During recordings, the oocytes were superfused using a Valvelink 16 (Automate Scientific) controller with a solution containing 40 mM Ba(OH)2, 50 mM NaOH, 1 mM KOH, 10 mM HEPES with the pH adjusted to pH 7.4 using HNO3. Two-electrode voltage-clamp experiments were performed using a GeneClamp 500B amplifier (Axon Instruments) controlled by a computer with a 1,200 MHz processor (Celeron, Gateway) using CLAMPEX 220.127.116.11 (Axon Instruments) and digitized at 1 kHz with a Digidata1332A (Axon Instruments). Electrodes were filled with 3 M KCl and had resistances of 0.1–0.3 MΩ. Leak currents were subtracted using a P/4 protocol. Ionic currents were analyzed with Clampfit 8.2 (Axon Instruments).
Supplementary Figure 1 Circular dichroism spectrum of CaV1.2 AID peptide in 150 mM KCl, 10 mM phosphate, 1 mM NaN3, pH 7.3, 15°C.
Supplementary Table 1 Y437 hydrogen bond network distances
We thank F. Findeisen for comments on the manuscript and for the ssCaVβ2a clone and M. Cho for expert molecular biology assistance. This work was supported by grants to DLM from NIH-NHLBI and American Heart Association and to FVP from the American Heart Association. DLM is an AHA Established Investigator.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.