Crystallographic basis for calcium regulation of sodium channels

The crystal structure of the Ca²⁺/CaM–Na_V1.5 DIII-IV linker complex was solved at 1.35 Å (Table S1). The asymmetric unit contains one CaM and one Na_V1.5 DIII-IV linker corresponding to residues 1491–1522. Fig. 1A shows the crystal structure of the complex with a single full-length CaM with four bound Ca²⁺ ions. A single amphipathic α-helix formed by residues 1489–1501 interacts with the C lobe through multiple contact points, with the remainder of the linker region (1502–1522) being intrinsically disordered, in agreement with previous studies (17). A total of ∼790 Ų (∼50%) is buried, of which ∼480 Ų is hydrophobic. The main anchor, Y1494, is completely buried in the CaM C lobe by interacting simultaneously with multiple CaM hydrophobic residues (Fig. 1 and Fig. S1A). Previously, Y1494 was identified as a residue of importance, as alanine substitution alters Ca²⁺/CaM binding energetics (as measured by ITC) and uncouples Ca²⁺ regulation of channel inactivation (9). The DIII-IV linker residue M1498 has a large contact area with CaM with 107 Ų out of 118 Ų buried (Fig. 1B), suggesting it could contribute to Ca²⁺/CaM binding and regulation of Na_Vs. Previous ITC experiments have shown that the interaction requires Ca²⁺ and is supported preferentially by the C lobe (K_d 19 μM versus ∼600 μM for the N lobe) (9). In addition, competition experiments whereby the C lobe is already bound to the DIII-IV linker show that the residual affinity of the N lobe is undetectable (Fig. 2C), suggesting the C lobe blocks access of the N lobe, as shown in other CaM complexes (18). The structure and ITC data therefore highlight the C lobe as the major and likely sole anchor point to the DIII-IV linker, and suggest that Ca²⁺/CaM may bridge different segments with the N lobe binding elsewhere, reminiscent of Ca²⁺/CaM bound to calcineurin (19) or Ca²⁺-activated K⁺ channels (20).

To date, a number of long Q-T type 3 (LQT3) cardiac disease mutations have been identified in the DIII-IV linker, and the positions of five of these can be directly observed in the crystal structure (Fig. 1B). M1498T, ΔK1500, and L1501V cause LQT3 (21–23), whereas Y1494N, ΔK1500, and G1502S (immediately adjacent to the structural interface between the DIII-IV linker and CaM) are involved in Brugada syndrome (22, 24, 25). Y1494, M1498, and L1501 are directly involved in the interaction with Ca²⁺/CaM, whereas the deletion of K1500 would affect the relative position of L1501 and all other downstream residues. G1502 is located at the C-terminal end of the α-helix, which marks the beginning of the intrinsically disordered region of the DIII-IV linker (Fig. 1B), but a serine at this site may interfere with the intrinsic, and apparent essential, disorder.

The energetic underpinnings of the Ca²⁺/CaM–DIII-IV linker interaction were determined by ITC, which can provide thermodynamic details of the interaction including affinity (K_a), enthalpy (ΔH), and entropy (ΔS) (26, 27). A sequence alignment demonstrates that the α-helical region of the DIII-IV linker is strictly conserved in all nine human Na_V isoforms (Fig. 2A). The crystallized construct, composed of residues 1491–1522, binds Ca²⁺/CaM in an exothermic fashion, much like the full-length DIII-IV linker but with higher affinity (Fig. 2B; K_d 0.35 μM vs. 2.98 μM for WT). The energetic impact of single Y1494A or Y1495A mutations on the WT or 1491–1522 constructs results in enthalpic penalties of 5 kcal/mol and 3 kcal/mol, respectively, which were matched with an equal gain in entropy, leaving the overall affinities unchanged (Table S2) (9). Given the identical energetic signature and enthalpy–entropy compensations, the interactions with Ca²⁺/CaM are likely to be the same in 1491–1522 and full-length DIII-IV linkers.

The DIII-IV linker is widely thought to form the inactivation gate of the channel, and introduced mutations at the triplet “IFM” motif have significant effects on sodium channel fast inactivation (14, 15). However, this motif has no role in Ca²⁺ regulation or CaM binding, as previously shown (9), and we show here that an outright deletion of the IFM motif does not impact CaM binding (Fig. 2D). Moreover, a peptide lacking the distal “FIF” motif formed by residues 1520–1522, previously suggested to be a Ca²⁺/CaM interaction site (10), bound the Ca²⁺/CaM with an affinity that corresponds well to that of the isolated C lobe of CaM (Fig. 2D and Table S2). A strictly conserved double-glutamate motif (E1489, E1490) is positioned close to CaM residue E142, an arrangement that could be responsible for decreasing the affinity of the full-length WT DIII-IV linker peptide. Consistent with this possibility, mutation of E1489 and E1490 to alanine produced full-length DIII-IV linkers with an approximately fivefold higher affinity than the wild-type peptide (Fig. 2D and Table S2). Interestingly, mutation of both E1489 and E1490 to lysine resulted in binding affinities similar to WT (Fig. 2D and Table S2), implying that the lowered affinity of the WT linker is due to steric clashes, rather than electrostatic repulsions.

Cytoplasmic Ca²⁺ modulates the steady-state inactivation of voltage-gated sodium channels (6, 7, 11, 28), an equilibrium relationship that provides a direct measure of channel availability at a given transmembrane potential (Fig. 3A). To better understand this process, DIII-IV linker residues involved in CaM binding, identified in the crystal structure and verified by ITC, were mutated in full-length Na_V1.5, and their contribution to Ca²⁺ regulation was investigated through patch-clamp electrophysiology. For instance, M1498 contributes significantly to the binding interface (Fig. 1B) with ∼90% of the residue buried, and the M1498A mutation strongly reduces both Ca²⁺/CaM and Ca²⁺/C lobe binding (K_d = 150 μM; Fig. 2D and Table S2) and ablates Ca²⁺ regulation (Fig. 3B; V_0.5 = −81.1 ± 0.3 and −82.5 ± 1.3 for 10 μM and 0 Ca²⁺, respectively). Alternatively, the E1489A/E1490A double mutant enhances Ca²⁺/CaM binding fivefold, and was engineered into full-length Na_V1.5 to produce “EE/AA” channels with significantly enhanced sensitivity to free Ca²⁺ compared with WT channels (Fig. 3C). Specifically, EE/AA channels responded to 300 nM free Ca²⁺ with a sizeable shift in the steady-state inactivation curve, whereas WT channels did not (P > 0.005). We next investigated the response to WT and EE/AA channels at multiple free Ca²⁺ concentrations in the recording pipette, measuring the size of the depolarizing shift in the steady-state inactivation curve (Fig. 3D). The data show that both wild-type and EE/AA channels respond to Ca²⁺ in a dose-dependent manner and display Hill coefficients greater than 1. Here the roughly fivefold increase in Ca²⁺/CaM binding found with ITC translated directly into a concomitant approximately fivefold increase in macroscopic sensitivity, with corresponding EC₅₀ values of 814 ± 69 nM and 147 ± 22 nM Ca²⁺ for WT and EE/AA channels, respectively (Table S5). The inverse correlation between Ca²⁺/C lobe affinity for the DIII-IV linker and inactivated state stability, that is, more binding results in less inactivation, points to a clear role for Ca²⁺/C lobe binding in the reduction of inactivation. Interestingly, the enhanced-affinity EE/AA channels, but not WT channels, displayed an apparently voltage-dependent slowing of inactivation only in the presence of Ca²⁺ (Fig. 3E and Table S6).

Ca²⁺ levels fluctuate during the excitation–contraction cycle in cardiac myocytes, and the total cytoplasmic Ca²⁺ concentration for half-maximal activation of contraction can reach ∼70 μM (2). Although it is known that the CaM–DIII-IV linker interaction requires Ca²⁺ for binding (9, 10), the dynamic range of Ca²⁺ levels, and more importantly the lower limit over which CaM can bind the inactivation gate, has remained unexplored. Here the data show that the interaction maintains affinity at low free Ca²⁺ levels (∼K_d of 25.3 μM in 100 nM free Ca²⁺) (Fig. 3F and Table S3). Thus, the interaction between Ca²⁺/CaM and the sodium channel DIII-IV linker can occur in a physiologically relevant range of the cardiac myocytes. Together, these experiments demonstrate that the observed Ca²⁺/CaM interaction in the crystal structure and the ITC characterizations are physiologically relevant, and that the DIII-IV linker is the final site of action for Ca²⁺/CaM regulation of channel inactivation.

The exact mechanism by which CaM might interact with the full-length C terminus has not been elucidated, but it is known that the IQ and EF-hand motifs undergo Ca²⁺-dependent conformational changes and can bind CaM in the presence and absence of Ca²⁺ (11, 29, 30). Data in Fig. 4A show that apo-CaM interacts with the IQ domain (residues 1896–1924) through its C lobe only, consistent with recent data (29). These interactions maintained the same profile and thermodynamic parameters with the C-terminal domain (CTD) spanning the EF hand and IQ domain (residues 1773–1925) (Table S4). In the presence of Ca²⁺, however, we found that both Ca²⁺/N lobe and Ca²⁺/C lobe can interact with the isolated IQ domain, with a binding preference for the Ca²⁺/N lobe (Table S4). Ca²⁺/CaM binds the CTD (residues 1773–1924) through a prerequisite interaction with the distal IQ domain (Fig. 4B), suggesting Na_Vs use a divergent mechanism from Ca_Vs, where the pre-IQ region can also bind CaM (31). The N value of the Ca²⁺/CaM–CTD interaction (∼0.5) suggests that one Ca²⁺/CaM can bind two CTDs. Indeed, both lobes are able to bind the CTD, with the affinity of the N lobe higher than for the C lobe (Fig. 4C). A competition experiment, whereby the Ca²⁺/C lobe is titrated into a preexisting mixture of CTD and Ca²⁺/N lobe, shows that the C lobe is no longer able to bind (Fig. 4C), showing that both lobes compete for an overlapping binding site. Given that the DIII-IV linker only shows significant binding to the Ca²⁺/C lobe and that the CTD can only bind one lobe at a given time, with a preference for the Ca²⁺/N lobe, these data support a simple model whereby Ca²⁺/CaM can bridge the DIII-IV linker and the CTD, with the C lobe bound to the inactivation gate.

All recombinant proteins were derived from cDNA from humans. The Na_V1.5 DIII-IV linker and C-terminal domain constructs were cloned into a modified PET28 (Novagen) vector (PET28HMT). Human CaM WT, N lobe (residues 1–78), and C lobe (residues 79–148) were generated previously (43). Proteins were purified as previously described (9), with additional details in SI Methods.

The Ca²⁺/CaM–DIII-IV domain complex was crystallized by hanging-drop vapor diffusion at 4 °C by mixing equal volumes of protein (10 mg/mL) and well solution containing 0.1 M MES (pH 6.5) and 50–60% (vol/vol) isopropanol. Additional details can be found in SI Methods.

ITC experiments were performed on an ITC-200 instrument (GE Healthcare). All samples were dialyzed for at least 24 h into buffer solution using a 3,500 molecular weight cutoff (MWCO) membrane for CaM and C-terminal proteins and a 1,000 MWCO membrane for peptides (Spectrum Laboratories). We assessed for affinities of CaM/DIII-IV linker/CTD in physiological solution as previously described (9). For the Ca²⁺ dependence studies, we prepared a solution containing 100 mM KCl, 10 mM Hepes, 20 mM EGTA, 20 mM CaCl₂. To obtain the desired free Ca²⁺ concentrations, additional EGTA or CaCl₂ was added, and the final free Ca²⁺ concentration was verified using a Ca²⁺-sensitive electrode (Denver Instrument). CaM concentration was held at 1 mM, with the DIII-IV peptide concentration at 100 μM. The binding isotherms were analyzed using a single-site binding model with the Microcal-modified version of Origin 7.0 (OriginLab).

Human Na_V1.5 (NM_198056) was a gift from Dr. Richard Horn and DIII-IV linker alanine mutations were generated using QuikChange site-directed mutagenesis (Stratagene). The calcium phosphate method (Invitrogen) was used to transiently cotransfect tSA-201 cells with channel DNA, eGFP, and CaM to minimize the possibility that overexpression of sodium channels could exhaust the endogenous pool of CaM. Details of the electrophysiology methodology can be found in SI Methods.