Unfolding of a temperature-sensitive domain controls voltage-gated channel activation

Abstract

Voltage-gated ion channels (VGICs) are outfitted with diverse cytoplasmic domains that impact function. To examine how such elements may affect VGIC behavior, we addressed how the bacterial voltage-gated sodium channel (BacNa_V) C-terminal cytoplasmic domain (CTD) affects function. Our studies show that the BacNa_V CTD exerts a profound influence on gating through a temperature-dependent unfolding transition in a discrete cytoplasmic domain, the neck domain, proximal to the pore. Structural and functional studies establish that the BacNa_V CTD comprises a bi-partite four-helix bundle that bears an unusual hydrophilic core whose integrity is central to the unfolding mechanism and that couples directly to the channel activation gate. Together, our findings define a general principle for how the widespread four-helix bundle cytoplasmic domain architecture can control VGIC responses, uncover a mechanism underlying the diverse BacNa_V voltage dependencies, and demonstrate that a discrete domain can encode the temperature dependent response of a channel.

Cells shape their electrical activity by controlling ion channel function in response to physical and chemical cues. Voltage-gated ion channels (VGICs) are exquisitely sensitive to transmembrane potential changes by virtue of a voltage-sensor domain that is embedded in the membrane bilayer (Vargas et al., 2012; Yu et al., 2005). Beyond the intrinsic ability to detect transmembrane voltage changes, VGIC superfamily members possess diverse intracellular domains (Yu et al., 2005) that are employed to tune voltage-dependent responses of a particular channel as a consequence of stimuli from signaling molecules (Morais-Cabral and Robertson, 2015; Yang et al., 2015). Although such domains provide a means for chemical cues to influence VGICs, the sensitivity to physical stimuli, such as temperature, in some VGIC superfamily members (Schneider et al., 2014; Vriens et al., 2014), has raised the question about whether there are equivalently specialized domains that can serve as temperature sensors (Bagriantsev et al., 2012; Brauchi et al., 2006; Grandl et al., 2008) or whether thermal responses arise from elements distributed throughout the channel (Chowdhury et al., 2014; Clapham and Miller, 2011). Moreover, despite notable advances in understanding the structures of some VGIC regulatory domains that respond to ligand or regulatory protein modulation (Jiang et al., 2002; Pioletti et al., 2006; Yuan et al., 2010; Zagotta et al., 2003), how conformational changes within these domains impact the channel pore and alter function remains incompletely understood.

Bacterial voltage-gated sodium channels (BacNa_Vs) share the six-transmembrane VGIC superfamily architecture and bear a four-helix bundle C-terminal cytoplasmic domain (CTD) that terminates in a four-stranded coiled-coil (Irie et al., 2012; Mio et al., 2010; Payandeh and Minor, 2015; Powl et al., 2010; Shaya et al., 2014). BacNa_Vs display remarkably diverse voltage responses (Payandeh and Minor, 2015; Scheuer, 2014), having activation potentials, V_1/2, that span a ~120 mV range (Scheuer, 2014), from −98 mV for Na_VAb from Arcobacter butzleri (Payandeh et al., 2012) to +27 mV for Na_VSp1 from Silicibacter pomeroyi (Shaya et al., 2014). The structural basis for this wide voltage response range is unknown (Scheuer, 2014). The CTD domain proximal to the channel pore termed the ‘neck’ is the most diverse element (Payandeh and Minor, 2015; Powl et al., 2010; Shaya et al., 2014) and appears to adopt varied degrees of structure in different BacNa_Vs (Bagneris et al., 2013; Shaya et al., 2014; Tsai et al., 2013). Studies of different BacNa_Vs indicate that the CTD is important for assembly (Bagneris et al., 2013; Mio et al., 2010; Powl et al., 2010; Tsai et al., 2013) and function (Bagneris et al., 2013; Irie et al., 2012; Shaya et al., 2014; Tsai et al., 2013). Nevertheless, a clear consensus for the mechanism by which the CTD influences channel function and the means by which it might influence the channel pore remains unknown.

Here, we show that the BacNa_V CTD has an integral role in controlling channel voltage-dependent behavior and that variation in neck composition and structure cause functional diversity. Moreover, we demonstrate that the neck has a temperature-dependent unfolding transition that is directly coupled to channel opening. These results establish that a discrete ion channel domain can serve as a temperature response element. The BacNa_V CTD location is shared with many VGIC superfamily member helical bundle domains (Howard et al., 2007; Paulsen et al., 2015; Uysal et al., 2009 ; Wiener et al., 2008; Yu et al., 2012), including channels not principally gated by voltage. This architectural commonality suggests that the principles uncovered here set a framework for understanding how four-helix bundle CTDs can modulate VGIC superfamily member function.

Results

BacNa_V cytosolic domain regulates voltage dependent channel opening

To probe BacNa_V CTD function, we focused on two BacNa_Vs having different activation signatures, Alkalilimnicola ehrlichii Na_VAe1 and Silicibacter pomeroyi Na_VSp1. Na_VSp1 displays robust voltage-dependent activity (Koishi et al., 2004; Shaya et al., 2014), whereas Na_VAe1 has not yielded functional data unless neck mutations are present (Shaya et al., 2014) (Figure 1A). We deleted the Na_VAe1 and Na_VSp1 CTDs at the hinge between the pore module and CTD (ending at Na_VAe1 His245 and Na_VSp1 His224, respectively (Figure S1A) (Shaya et al., 2014)). Contrary to the loss of function reported for other BacNa_V CTD deletions (Mio et al., 2010; Tsai et al., 2013), Na_VAe1Δ_CTD and Na_VSp1Δ_CTD produced robust voltage dependent currents, demonstrating that CTD removal did not strongly affect folding or tetramerization (Figure 1A). However, CTD deletion had robust effects on voltage dependent activation, causing large left shifts in V_1/2 relative to the references (V_1/2 = 19.1 ± 1.9 and 32.1 ± 1.1 mV, for Na_VAe1Δ_CTD and a functional mutant bearing a neck triple glycine substitution Na_VAe1_GGG (Shaya et al., 2014), respectively, and −7.4. ± 2.8 and 28.9 ± 1.3, for Na_VSp1Δ_CTD and Na_VSp1, respectively (Figure 1G–H and Table S1)). Thus, CTD deletion produces Na_VAe1 and Na_VSp1 channels that are more readily opened by voltage.

Figure 1. BacNa_V CTD affects function, Related to Figure S1 and Table S1.

Functional comparison of: A, Na_VAe1, Na_VSp1, Na_VAe1Δ_CTD, and Na_VSp1Δ_CTD; B, Na_VAe1_SpCTD and Na_VSp1_AeCTD; C, Na_VAe1_SpNeck and Na_VSp1_AeCC; and D, Na_VAe1Δ_CC and Na_VSp1Δ_CC. Cartoons depict two BacNa_V subunits. Voltage-sensor domain (VSD), pore domain (PD), Neck, and Coiled-coil (CC) are labeled. Na_VAe1 and Na_VSp1 elements are blue and orange, respectively. E, Voltage protocols for Na_VAe1 (top), Na_VSp1 (bottom), and chimeras. F, Na_VSp1_AeCTD/GGG exemplar currents. G-H, Voltage-dependent activation curves for: G, Na_VAe1, Na_VAe1_SpCTD, Na_VAe1_SpNeck, Na_VAe1Δ_CTD, and Na_VAe1Δ_CC, and H, Na_VSp1, Na_VSp1_AeCC, Na_VSp1Δ_CTD, and Na_VSp1Δ_CC. I, Western blot of the total lysate and surface-biotinylated fraction for the indicated constructs probed using the specified antibodies. See also Figure S1.

The CTD has two parts, the neck (Payandeh and Minor, 2015; Shaya et al., 2014), proximal to the transmembrane pore, and a C-terminal four stranded coiled-coil (Irie et al., 2012; Mio et al., 2010; Payandeh and Minor, 2015; Powl et al., 2010; Shaya et al., 2014). Because CTD deletion had such a strong effect on Na_VAe1 and Na_VSp1 function, we created a series of chimeras (Figure S1A) to define whether the effects caused by the CTD arise from the neck, coiled-coil, or both. CTD exchange between Na_VAe1 and Na_VSp1 yielded channels, Na_VAe1_SpCTD and Na_VSp1_AeCTD, that phenocopied the properties of the CTD parent; Na_VAe1_SpCTD was functional (V_1/2 = 38.8 ± 2.5 mV) whereas Na_VSp1_AeCTD showed no activity (Figure 1B and Table S1). Exchanging the Na_VAe1 coiled-coil into Na_VAe1_SpCTD to make an Na_VAe1 chimera bearing the Na_VSp1 neck and Na_VAe1 coiled-coil, Na_VAe1_SpNeck, produced a channel having voltage-dependent gating that was essentially identical to Na_VAe1_SpCTD (V_1/2 = 32.7 ± 2.0 and 38.8 ± 2.5 mV for Na_VAe1_SpNeck and Na_VAe1_SpCTD, respectively) (Figures 1C and 1G, Table S1). Substitution of the Na_VSp1 neck into Na_VSp1_AeCTD produced a Na_VSp1 chimera having the Na_VAe1 coiled-coil, Na_VSp1_AeCC that displayed voltage-dependent gating identical to Na_VSp1 (V_1/2 = 26.1 ± 1.0 and 28.9 ± 1.3 mV, for Na_VSp1_AeCC and Na_VSp1 respectively) (Figures 1C and 1H, and Table S1). Taken together, these results demonstrate that coiled-coil identity has minimal functional effects and that the neck is the principal source of the differences in Na_VAe1 and Na_VSp1 voltage-dependent behaviors. Although the coiled-coil did not influence channel voltage-dependent properties, it seemed possible that its ability to constrain the neck C-terminal end might be important for the neck to affect gating. Therefore, we examined Na_VAe1 and Na_VSp1 mutants in which the neck was intact but the coiled-coil was deleted, Na_VAe1Δ_CC and Na_VSp1Δ_CC. Both yielded channels having a V_1/2 indistinguishable from complete CTD deletion (19.1 ± 1.9 and 17.4 ± 1.7 mV for Na_VAe1_ΔCC and Na_VAe1_ΔCTD; −7.4 ± 2.8 and −9.7 ±1.9 mV for Na_VSp1_ΔCC and Na_VSp1_ΔCC, respectively) (Figures 1D, G and H, Table S1). Thus, together with the chimera results, these data indicate that the coiled-coil is required to constrain the C-terminal end of the neck, although its identity has minimal effect on function.

Because we were unable to measure currents from Na_VAe1 and Na_VSp1_AeCTD (Figures 1A–B), channels that have a wild-type Na_VAe1 neck, we tested whether these proteins had plasma membrane expression to resolve whether the lack of activity came from an absence of surface expression or from channels that could not be opened. We placed an N-terminal hemaggluttinin (HA) tag on Na_VAe1, Na_VSp1_AeCTD, and Na_VSp1 and used a surface biotinylation assay to assess plasma membrane expression. Streptavidin capture of the surface-biotinylated fraction followed by anti-HA antibody detection showed clear signals for all three channels. By contrast, this fraction had no signal for the intracellular control, green fluorescent protein, GFP, when probed with an anti-GFP antibody. Thus, the non-functional channels, Na_VAe1 and Na_VSp1_AeCTD, are indeed expressed on the plasma membrane (Figure 1I). Moreover, similar to the effects on Na_VAe1 (Shaya et al., 2014), inclusion of a triple-glycine substitution in the neck domain of Na_VSp1_AeCTD generated a functional channel (Na_VSp1_AeCTD/GGG V_1/2= 5.75 ± 2.6 mV, Figures 1F and H, Table S1). Together with the results of the chimeras and previous neck mutants (Shaya et al., 2014), these findings suggest that Na_VAe1 and Na_VSp1_AeCTD fail to produce currents because the Na_VAe1 neck produces a V_1/2 that is outside of the measureable voltage range of our experiments.

Our data demonstrate that the CTD functions as a regulatory module whose intrinsic properties enable it to exert diverse effects on BacNa_V voltage-dependent gating. The neck influence is strong on voltage-dependent activation and contrasts with its modest effects on voltage-dependent inactivation (Figures 1 and S1B–C, Table S1), a process for which the pore domain is important (Pavlov et al., 2005; Payandeh et al., 2012; Shaya et al., 2014). Thus, our investigations establish that the origins of the dramatic modulation of voltage-dependent gating reside in the neck and require the neck to be constrained at the C-terminal end by the coiled-coil domain.

Diverse BacNa_Vs CTDs alter Na_VSp1 transmembrane domain voltage responses

Prior functional studies reported BacNa_V V_1/2 activation values spanning a ~120 mV range (Koishi et al., 2004; Payandeh et al., 2012; Ren et al., 2001; Shaya et al., 2014; Ulmschneider et al., 2013). The source of this diversity has been unclear (Scheuer, 2014). Given the profound effects of the CTD on Na_VSp1 and Na_VAe1, we next asked how replacement of the Na_VSp1 CTD with CTDs from previously studied BacNa_Vs having diverse lengths and compositions would affect function: Na_VBh1 (Na_VSp1_BhCTD), Na_VMs (Na_VSp1_MsCTD), Na_VAb (Na_VSp1_AbCTD), Na_VAb1 (Shaya et al., 2011) (Na_VSp1_Ab1CTD), and Na_VPz (Na_VSp1_PzCTD) (Figures 2A and S1A). In whole cell recordings each of these channels, except Na_VSp1_AbCTD, was functional (Figure 2B). Strikingly, the CTD substitutions had diverse effects ranging from little change in voltage-dependent activation relative to Na_VSp1 for the Na_VPz CTD (ΔV_1/2 = −4.2 mV) to producing large voltage-dependent activation shifts in both the negative (ΔV_1/2 = −44.5 ± 4.4 and −27.3 ± 3.8 mV, Na_VBh1 and Na_VMs CTDs, respectively) and positive directions (ΔV_1/2 = +20.4 ± 5.8 mV Na_VAb1 CTD) (Figure 2C, Table S1). For the Na_VSp1_BhCTD the activation threshold was similar to that of Na_VSp1Δ_CTD (ΔV_1/2 = −44.5 ± 4.4 mV and −36.3 ± 5.4 mV, respectively), a result that is in line with data suggesting that the Na_VBh1 neck is disordered (Powl et al., 2010) and the observation that Na_VBh1 coiled-coil deletion results in channels having voltage-dependent activation indistinguishable from wild type (Mio et al., 2010). By contrast, the CTD swaps had a more modest effect on voltage-dependent inactivation (Figure S1D, Table S1).

Figure 2. CTD chimeras alter Na_VSp1 voltage dependence, Related to Figures S1 and S2 and Table S1 and Table S2.

A, Sequence alignment of Na_VAe1, Alkalimnicola erlichii (Shaya et al., 2014; Shaya et al., 2011), Na_VSp1, Silicibacter pomeroyi (Koishi et al., 2004; Shaya et al., 2011); Na_VAb1, Alcanivorax borkumensis (Shaya et al., 2011); Na_VBh1 (NaChBac), Bacillus halodurans (Ren et al., 2001); Na_VAb, Arcobacter butzleri (Payandeh et al., 2011); Na_VMs, Magnetococcus sp. (McCusker et al., 2012); and Na_VPz, Paracoccus zeaxanthinifaciens (Koishi et al., 2004) indicated regions. S6, Neck, Coiled-coil ‘a’–‘d’ positions, and 3G mutation site are indicated. B, NaSp1, Na_VSp1_BhCTD, Na_VSp1_MsCTD, Na_VSp1_Ab1CTD, Na_VSp1_PzCTD, and Na_VSp1_AbCTD exemplar currents and voltage protocol. Cartoons depict two channel subunits. C, Activation curves for ‘B’. D, Temperature dependence of Na_VSp1 activation. E, V_1/2 temperature dependence for the indicated channels. Lines show linear fit.

The varied effects of the different CTD chimeras on the voltage-dependence of the Na_VSp1 transmembrane core demonstrate that each BacNa_V CTD has a distinct effect on channel activation (Figure 2B) and that a simple BacNa_V CTD transplant can tune the voltage-dependence of activation of the transmembrane core over a wide range, ~65 mV (Figure 2C). These results, taken together with the experiments indicating the clear involvement of the BacNa_V neck in channel function (Figure 1) and lack of sensitivity to coiled-coil identity (Figure 1C) strongly support a critical role for the neck in BacNa_V gating. Although specific elements of the BacNa_V transmembrane domains, such as the voltage sensors, must set some of the channel voltage-dependent properties, our findings together with the observation that the neck is the most variable BacNa_V feature (Payandeh and Minor, 2015), indicate that much of the observed variety in BacNa_V voltage-dependences originates from the neck region. These domains display varied degrees of structure from disordered (Bagneris et al., 2013; Powl et al., 2010) to completely ordered (Shaya et al., 2014). Thus, our data points to a common mechanism for tuning voltage dependencies within the BacNa_V family based on the ability of the neck to adopt structure.

Temperature-dependent change in the BacNa_V neck affects gating

Previous studies suggested that the neck undergoes an order → disorder transition during channel opening (Shaya et al., 2014) and pose a hypothesis that predicts very different temperature dependences for BacNa_Vs in which the neck is stably folded structure versus those in which it is disordered. To test this, we first examined whether Na_VSp1 voltage-dependent activation was temperature dependent and whether this response could be influenced by neck properties. The V_1/2 of Na_VSp1 activation showed a clear temperature dependence, moving ~35 mV in the hyperpolarized direction as temperature increased from 18° to 35°C (Figures 2D–E, and Table S2). This response was eliminated by increasing the neck flexibility with a triple glycine substitution, Na_VSp1_GGG (Shaya et al., 2014), or by CTD deletion, Na_VSp1Δ_CTD (Figure 2E, Table S2), supporting the idea that neck structure is crucial for temperature-dependent gating changes.

Because the Na_VSp1 CTD chimeras displayed diverse activation V_1/2 values (Figure 2C), we next probed whether these CTD substitutions affected channel temperature responses. The chimeras having CTDs bearing disordered necks, Na_VSp1_BhCTD and Na_VSp1_MsCTD (Bagneris et al., 2013; Powl et al., 2010), yielded channels that lacked a temperature response in activation V_1/2 (Figure 2E, Table S2). By contrast, the activation V_1/2 of Na_VSp1_PzCTD and Na_VSp1_Ab1CTD showed clear temperature dependence. This response was similar in magnitude to Na_VSp1 (Figure 2E) and shows that the role of the CTD in setting channel temperature-dependent properties is general.

Finally, we examined whether the V_1/2 changes caused by the CTD substitutions were related to neck structure by examining the consequences of neck triple glycine mutations in Na_VSp1_MsCTD, Na_VSp1_Ab1CTD, and Na_VSp1_PzCTD (Figure S2A–B). Neck disruption had no effect on the activation V_1/2 for the chimera having a disordered neck, Na_VSp1_MsCTD (ΔV_1/2 = 0.9 mV Na_VSp1_MsCTD/GGG relative to Na_VSp1_MsCTD Figure S2C, Table S1). However, for both chimeras having a temperature dependent V_1/2 similar to Na_VSp1, Na_VSp1_PzCTD and Na_VSp1_Ab1CTD, the GGG mutation caused a large activation V_1/2 large left-shift (ΔV_1/2 = −39.3 and −47.2 mV respectively for Na_VSp1_PzCTD and Na_VSp1_Ab1CTD, relative to the parent chimeras, Figure S2D and S2E, Table S1). This magnitude change is similar to that in Na_VSp1_GGG (ΔV_1/2 = −39.4 mV, Fig. S2F, Table S1). Given that the triple glycine mutation eliminates Na_VSp1 temperature dependence, these data suggest that the modulatory effects on V_1/2 and the temperature dependent gating properties in Na_VSp1_PzCTD and Na_VSp1_Ab1CTD arise from neck domain order. Taken together, our data strongly support the hypothesis that a protein unfolding transition in the neck domain is coupled to channel opening and demonstrate that a discrete channel domain can act as a temperature sensor.

Structures of Na_VAe1p neck mutants reveal disordered neck and closed pore

To define the structural consequences of polyglycine mutations intended to disrupt the neck, we determined crystal structures of two neck domain mutants of ‘pore-only’ channel Na_VAe1p (Shaya et al., 2014), Na_vAe1p-3G and Na_vAe1p-7G, at 3.70 Å and 3.80 Å resolution, respectively (Figures 3A–D, Supplementary Figure S3 and Table S3). Molecular replacement using the Na_vAe1p transmembrane portion (Shaya et al., 2014) yielded maps having well-defined electron density for the transmembrane regions and part of the CTDs (Figures S4A and S4B). Model building and refinement showed traceable density for residues that comprise the neck C-terminal ends and complete coiled-coils (Na_vAe1p-3G Ile254-Arg283 and Na_vAe1p-7G Glu257-Arg283)(Figures 3A and B) but lacked density for the polyglycines and residues that frame these substitutions (Na_vAe1p-3G Ser243-Arg253 and Na_vAe1p-7G Ser243-Gln256). This localized loss of structure confirms the increase in neck flexibility and shows that the disruption propagates beyond the polyglycines (Figure 3D).

Figure 3. BacNa_V neck disruption structural outcomes, Related to Figures S3 and S4, and Table S3.

Cartoons of two subunits of A, Na_VAe1p-3G (dark blue) and B, Na_VAe1p-7G (firebrick) structures. Grey dashes indicate regions lacking electron density. Residues defining the electron density limits are indicated. Selectivity filter is orange. C, Superposition of Na_VAe1p (orange) (Shaya et al., 2014), Na_VAe1p-3G (dark blue), and Na_VAe1p-7G (firebrick). Angle shows Na_VAe1p-7G coiled-coil displacement. Selectivity filter ion positions are shown. D, S6-CTD sequences for the indicated channels. Polyglycine positions are red. Regions lacking electron density are grey. Coiled-coil ‘a–d’ repeat is orange. E, and F, CD spectra for the indicated proteins at 4°C. G, Thermal denaturation curves for the indicated proteins.

Although the Na_vAe1p-3G and Na_vAe1p-7G necks lack structure, both CTD coiled-coils remained intact. These were displaced from the channel central axis by ~28° (Figure 3C) and in agreement with its increased disorder, this displacement was larger for Na_vAe1p-7G and included a ~8Å movement towards the pore (Figure 3C). Once the neck is disrupted, the pore domain and coiled-coil appear to be independent. We anticipate that free from the crystal lattice constraints, the coiled-coil could move independently relative to the pore as solution studies suggest (Bagneris et al., 2013). The fact that the neck disruption does not propagate into the coiled-coil agrees with our observation that coiled-coil identity is not crucial for function but that its presence is essential for the neck to affect function (Figures 1C–D).

Although 3G and 7G substitutions make full length BacNa_Vs easier to open, Na_vAe1p-3G and Na_vAe1p-7G pore domains are closed and match the Na_VAe1p closed structure (Figure S4C) (Shaya et al., 2014) (RMSD_Cα of 0.67 Å and 0.73 Å, respectively). Assignment of a closed conformation is further supported by anomalous difference maps of selenomethionine substituted Na_vAe1p-3G (Figure S4D and Table S3) that reveal clear density on the pore domain central axis for the seleniums of the activation gate residue, Met241 (Shaya et al., 2014). Na_VAe1p-3G, SeMet Na_VAe1p-3G, and Na_VAe1p-7G structures also revealed density in the selectivity filter ‘site 2’ position (Tang et al., 2014) (Figures S4E–G) that we modeled as a calcium atom based on anomalous density (Figure S4F) and the presence of calcium in the crystallization conditions.

As there is no structure of Na_vSp1 or the ‘pore–only’ Na_VSp1p (Shaya et al., 2011), we turned to circular dichroism (CD) spectroscopy to test whether the Na_VAe1p-3G and Na_VAe1p-7G structural changes had parallels in Na_vSp1p. Na_VAe1p, Na_VAe1p-3G, and Na_VAe1p-7G CD spectra showed classic helical features of minima at 208 and 222 nm (Berova et al., 2000). Notably, the intensity of these minima was reduced in Na_VAe1p-3G and Na_VAe1p-7G in an order that matched the crystal structures and was reduced farther by CTD deletion in Na_VAe1p-ΔCTD (Figure 3E). Comparison of the CD spectra of Na_vSp1p (Shaya et al., 2011) with Na_vSp1p-3G caused a similar loss of helical structure (Figures 3D and 3F). To test whether neck disruption affected the thermal stability, we measured the temperature dependence of the CD signal at 222 nm. The data show that in Na_VAe1p introduction of neck region glycines increases the thermal liability of the measured transition following the rank order as expected from the crystal structures (Na_VAe1p > Na_VAep1-3G > Na_VAe1p-7G) (Figure 3G). Notably, Na_VAe1p-ΔCTD lacks a cooperative thermal transition, demonstrating that the measured changes in thermal behavior arise from the CTD (Figure 3G). Further, Na_VSp1p is less stable than Na_VAe1p and the introduction of the triple glycine neck mutant in Na_VSp1p eliminated the thermal transition, in agreement with the function of the parent channels (Figures 1 and 2). Together, the crystallographic and CD data demonstrate that increasing BacNa_V neck flexibility leads to a loss of structure that is restricted to the neck, show that these changes increase the neck thermal sensitivity, and indicate that this disruption is the source of the functional changes caused by polyglycine substitution.

EPR spectroscopy reveals changes in neck dynamics

To investigate the CTD domain dynamics, we used site specific spin labeling with the nitroxide spin probe (1-Oxyl-2,2,5,5-tetramethylpyrroline-3-methyl) methanethiosulfonate (MTSSL) and electron paramagnetic resonance (EPR) studies (McHaourab et al., 2011). We probed select positions on the exterior of the Na_VAe1p and Na_VAe1p-3G CTDs by labeling three neck sites (Glu247 and Ala251 that frame polyglycine sites in Na_VAe1p-3G, and Glu255 at the beginning of the Na_VAe1p-3G ordered region)(Figures 3A and 4A), three coiled-coil sites (Glu266, Gln269, and Asp273), and a site at the protein C-terminus (Arg283) (Figure 4A). Double electron-electron resonance (DEER) spectroscopy indicated that, with the exception of Arg283, the probe positions in Na_VAe1p have spin echo decays with a well-defined oscillation (Figure 4B). The corresponding distance distributions match those expected from the structure (~25Å and ~35Å for adjacent and diagonal nitroxide positions, respectively). By contrast, introduction of triple glycine at residues 248–250 (Figure 4A) caused clear changes at Glu247 and Ala251. These decays showed evidence of superimposed longer component that is manifested in the distance distributions. We interpret this component as indicative of increased disorder (Figure 4B). Notably, minimal changes occurred at Glu255, in complete agreement with the order seen at this site in the crystal structure (Figure 3A). Further, no differences were observed at any coiled-coil positions. Thus, these data reinforce the view from our crystallographic studies that introduction of polyglycine sequences into the neck increase its mobility and spare the coiled-coil structure.

Figure 4. Na_VAe1p EPR studies, Related to Figure S5.

A, S6-CTD region sequences. Spin label positions, polyglycine substitutions and, Coiled-coil ‘a–d’ repeat are highlighted green, red, and orange, respectively. Na_VAe1p-3G residues lacking electron density are grey. B, Na_VAe1p cartoon of two subunits and spin label sites. Pore domain, Neck and Coiled-coil are slate, sand, and orange, respectively, Red and black boxes denote Neck and coiled-coil spin-label positions, respectively. Panels show DEER decays, distance distributions, and CW spectra. Asterisks indicate the new Na_VAe1p-3G long-range distances.

There are clear functional differences between Na_VAe1 and Na_VSp1 that originate in the CTD (Figure 1). Therefore, we probed Na_VSp1p at positions equivalent to those tested in Na_VAe1p and Na_VAe1p-3G (Figure S5) to see whether there was a structural correlate underlying the diverse functional properties. Na_VSp1p continuous wave (CW) and DEER studies reveal notable differences in the dynamics of the neck versus the coiled-coil. By contrast to Na_VAe1p, the Na_VSp1p neck positions consistently display DEER decays that have a longer distance component, similar to those in Na_VAe1p-3G, whereas the Na_VSp1p coiled-coiled DEER signals resemble what is observed in Na_VAe1p (Figure S5B). This differences between EPR probe mobility in the neck and coiled-coil is similar to those observed in Na_VMs (Bagneris et al., 2013). Importantly, our data strongly support the notion from the chimera (Figure 2) and CD experiments (Figure 3G) that the neck domains of different BacNa_Vs have diverse degrees of inherent order and that these differences are the origins of functional diversity.

Na_VAe1p neck has a hydrophilic core and a π-stack structure important for gating

In parallel with our studies of Na_VAe1p mutants, we obtained Na_VAe1p crystals that diffracted to superior resolution than previously reported (Shaya et al., 2014) (2.95Å vs. 3.46Å, respectively) (Figures S6A and S6B and Table S3). Although the overall architecture is unchanged (Figure S6C), the increased resolution revealed previously uncharacterized CTD features. Notably, there is a striking bipartite organization in the core of the CTD neck and coiled-coil domains. The coiled-coil core follows the classic heptad repeat ‘a-d’ packing (Lupas and Gruber, 2005) (Figures 5A–C) and comprises five layers of hydrophobic residues (Met267, Ile271, Leu274, Val278, and Leu281) and a terminal, more open layer at Ser285 (Figure 5B–C). By contrast, the neck core is composed almost exclusively of hydrophilic sidechains. These also follow a heptad pattern but include a one residue skip at the Gln256-Gln257 junction (Figure 5A). Consistent with the splaying of the neck helices (Shaya et al., 2014), the neck has a less well-packed core than the coiled-coil (Figure 5B–C). The stark differences between the structures and compositions of the neck and coiled-coil cores suggest that the neck is metastable and provide an explanation for why it is able to change structure during channel gating.

Figure 5. Na_VAe1p CTD structure, Related to Figures S6 and S7 and Table S3.

A, Na_VAe1 sequence. Neck and Coiled-coil ‘a’–‘d’ core residues, His245 and Glu256 are highlighted. B, Na_VAe1p CTD cartoon showing core residues. Neck ion, ${\text{Cl}}_{\mathrm{N}}^{-}$, and Coiled-coil ion, ${\text{Cl}}_{\text{CC}}^{-}$ are indicated. S6, Neck and Coiled-coil are slate, sand, and orange, respectively C, Na_VAe1p CTD π-stack, ‘a’ and ‘d’ layers, and Coiled-coil ion site packing geometries. D–E, π-stack ion binding site D, Side and E, top views. His245, Trp246, and Neck ion, ${\text{Cl}}_{\mathrm{N}}^{-}$, are indicated F, Coiled–coil ion binding site. Glu257, Ala260, Arg264, and Coiled-coil ion, ${\text{Cl}}_{\text{CC}}^{-}$ are indicated. F, Na_VAe1_H245G, Na_VAe1_W246G, and Na_VAe1_H245G/W246G exemplar currents and voltage-dependence of activation and inactivation. Colors are as in Figure 4.

The 2.95 Å resolution Na_VAe1p structure revealed a number of interesting non-protein entities associated with the channel. There are two ions in the CTD core. One is at the N-terminal end of the neck, the ‘neck ion’ (Shaya et al., 2014), and a second is at the neck/coiled-coil junction, the ‘coiled-coil ion’ (Figure 5B). Data collected at the bromine absorption edge from Na_VAe1p crystals grown in 200 mM NaBr and soaked in cryoprotectant solutions containing 200 mM or 0 mM NaBr showed strong anomalous densities for both ions in 200 mM NaBr, but only for the coiled-coil ion in 0 mM NaBr (Figure S6D), demonstrating that both CTD ions are halogens and that the neck ion is labile. Based on these observations and the presence of chloride in the 2.95 Å structure crystallization conditions, we modeled both ions as chloride (Figure S6E). The selectivity filter has density corresponding to two sodium ions bridged by a water molecule. One sodium ion occupies the level of the Glu197 sidechain (corresponding to ‘site 2’ (Tang et al., 2014)), and the second is found at the level of Thr195 (‘site 3’) (Figure S6F). We also identified a number of lipids including one occupying a site observed in a number of other BacNa_V structures (Payandeh and Minor, 2015) that is located next to the P1 pore helix and wedged between the subunits (Figure S6G).

The neck ion binding site arises from an unusual motif, the ‘π-stack’, in which Trp246 from one subunit makes a face-to-face stacking interaction with His245 from the neighbor (Figures 5B–E) and creates a neck ion binding site coordinated by the Trp246 indole nitrogens. The coiled-coil ion binding site comprises successive ‘a-d-a’ positions in the CTD core (Figures 5B, 5C, and 5F). Four guanido moieties from Arg264 that occupy the first ‘a’ position of the coiled-coil ‘a–d’ repeat coordinate the halide ion (Figure 5F). This ionic complex forms an ‘electrostatic pin’ in the helical bundle core. The Arg264 sidechains extend into the space that should be occupied by the preceding ‘d’ sidechains of the neck core. The presence of a small residue at this position, Ala260, accommodates the Arg264 guanido moieties and permits them to interact with the Glu257 carboxylates from the ‘a’ position above (Figures 5C and 5F). Interestingly, the ‘electrostatic pin’ formed by the coiled-coil ion binding site marks where the CTD helices diverge from the canonical coiled-coil packing. Further, in the Na_VAe1p-3G and Na_VAe1p-7G structures, disruption of the helical structure propagates through the π-stack but stops at the coiled-coil ion binding site (Figure S6E). These observations agree with the apparent neck and coiled-coil ions binding differences and the metastable nature of the neck hydrophilic core versus the stable coiled-coil hydrophobic core. π-stack and coiled-coil ion motifs signatures occur in other BacNa_V sequences and indicate that these structures are present in other channels (Figure S7).

Because the Na_VAe1p-3G and Na_VAe1p-7G structures showed disrupted π-stack elements due to the polyglycine mutations (Figures 3A–B, 3D, and S6E), we wanted to test whether direct disruption of this structural element would affect function. As with other Na_VAe1 neck destabilizing mutations (Shaya et al., 2014), mutation of either π-stack residue to glycine resulted in measurable currents (Figure 5G). As expected from the intersubunit stacking Na_VAe1_H245G/W246G had essentially identical voltage-dependent activation and inactivation properties compared to Na_VAe1_H245G and Na_VAe1_W246G, demonstrating that the two π-stack elements are co-dependent (Figure 5G, Table S1). A previous Na_VAe1p_H245G low resolution structure showed that π-stack disruption leads to neck ion loss the but little perturbation to the neck helix (Shaya et al., 2014). Thus, theπ-stack architecture serves as a type of lock in the Na_VAe1 neck to oppose channel opening.

BacNa_V neck is energetically coupled to the channel gate

Does the BacNa_V neck affect function by impacting the channel gate, the voltage sensors, or both? To answer this question, we pursued a double mutant cycle strategy (DeCaen et al., 2009; Yifrach and MacKinnon, 2002) examining whether Na_VSp1 neck by the GGG substitution, Na_VSp1_GGG, had additive or non-additive effects on voltage gating when placed in the context of mutations that perturb the gate or the voltage sensors. To test whether the neck and gate are coupled, we compared the properties of the Na_VSp1 channel gate mutant, Na_VSp1_M220A (Shaya et al., 2014), Na_VSp1_GGG, and a channel bearing both mutations Na_VSp1_M220A/GGG (Figures 6A–B). To probe for coupling to the voltage sensors, we neutralized each of the four arginines at BacNa_V S4 voltage sensor positions R1–R4 (Payandeh et al., 2011; Zhang et al., 2012) by mutation to glutamine (Figure 6C) in order to perturb the ease of voltage sensor movement and measured the properties of Na_VSp1_R1Q, Na_VSp1_R2Q, Na_VSp1_R3Q, and Na_VSp1_R4Q, alone and in combination with the GGG neck mutation (Figures 6D–E).

Figure 6. BacNa_V CTD couples to the intracellular gate, Related to Table S4.

A, Na_VSp1, Na_VSp1_GGG, Na_VSp1_M220A, and Na_VSp1_M220A/GGG exemplar currents. B, Activation curves from ‘A’. C, Na_VSp1 S4 mutant sites. D, Na_VSp1 R1–R4 and R1–R4/GGG mutant exemplar currents. E, Voltage-dependent activation curves for the indicated mutants. Na_VSp1 (tan dashes) and Na_VSp1_GGG (black dashes) curves are shown for reference.

As observed previously, gate destabilization, Na_VSp1_M220A (Shaya et al., 2014), and neck disruption, Na_VSp1_GGG (Shaya et al., 2014), caused substantial leftward shifts in the activation V_1/2 (27.7 ± 1.6 mV, −10.5 ± 2.1 mV, and −18.6 ± 2.7 mV for V_1/2 of Na_VSp1, Na_VSp1_GGG, and Na_VSp1_M220A, respectively, Figure 6B, Table S4). Combining both, Na_VSp1_M220A/GGG, resulted in a V_1/2 left shift beyond that of the individual changes (−28.3 ± 4.0 mV, Figure 6B and Table S4). R1 neutralization did not affect the activation V_1/2 (31.1 ± 1.2 mV), whereas other arginine position neutralizations shifted the activation V_1/2 in either the positive (Na_VSp1_R2Q, and Na_VSp1_R4Q, 44.8 ± 2.1 mV and 59.9 ± 3.2 mV, respectively) or negative (Na_VSp1_R3Q, 11.7 ± 2.3 mV) directions relative to Na_VSp1 (Figure 6E and Table S4). In each case, combination with GGG neck mutation caused a substantial left shift in V_1/2 (Figure 6E) (Na_VSp1_R1Q/GGG, Na_VSp1_R2Q/GGG, Na_VSp1_R3Q/GGG, and Na_VSp1_R4Q/GGG, −9.2 ± 1.9 mV, 44.8 ± 2.1 mV, −24.5± 1.8 mV and 32.7± 2.0 mV, respectively).

By taking changes in V_1/2, estimated gating charge, Z(e₀), and slope factor (k) (DeCaen et al., 2009; Yifrach and MacKinnon, 2002), we determined the activation free energy at 0 mV. Comparison of measured double mutant free energy perturbations relative to Na_VSp1 (ΔΔG°_obs) with values calculated additive effects of individual mutants (ΔΔG°_calc) revealed a clear contrast between how neck perturbation affects the pore and voltage sensors (Table S4). ΔΔG°_obs for the neck-gate combination, Na_VSp1_M220A/GGG, −3.1 kcal mol⁻¹, is substantially less than that expected from the individual (ΔΔG°_calc= −5.0 kcal mol⁻¹) (Table S4). By contrast, ΔΔG°_obs =ΔΔG°_calc for all neck-voltage sensor pairs (Table S4). Thus, these data indicate that the neck affects gating voltage dependence by direct perturbation of the channel gate. Given its location and the importance of neck structure, such an effect must come from the ability of the neck to constrain pore opening.

Discussion

VGIC cytosolic domains are important channel modulation loci. Hence, there is a great interest in understanding how their structural transformations affect VGIC transmembrane channel core. The simple BacNa_V CTD architecture, comprising four parallel helices, provides an elegant paradigm for defining how VGIC CTDs can influence channel gating. Our studies of BacNa_V CTDs reveal a VGIC modulation mechanism built on the bipartite architecture of the BacNa_V CTD four-helix bundle. The key feature, the neck domain, is a membrane proximal four-helix bundle bearing a hydrophilic core that forms a metastable structure. This domain is constrained on its N- and C-termini by the channel pore domain and a classic parallel four-stranded coiled-coil, respectively. These physical constraints are essential for the neck to influence voltage-dependent gating and their importance is supported by two key observations: 1) Deletions of the entire CTD (Figure 1G and H, Table S1) shift the voltage-dependent gating in the hyperpolarized direction in a manner equivalent to deletion of only the coiled-coil (Figure 1, Table S1) and 2) mutant cycle analysis shows strong energetic coupling between the activation gate of the channel pore and the neck (Figure 6, Table S4).

The origin of the wide range of BacNa_V voltage-dependent activation responses has been unclear (Scheuer, 2014). The diverse and largely hydrophilic nature of BacNa_V neck domains contrasts with the well-conserved CTD coiled-coil and high conservation of key voltage sensor domain elements (Payandeh and Minor, 2015). This neck sequence diversity, together with the capacity of different BacNa_V CTDs to tune voltage-responses of the Na_VSp1 transmembrane core by >65 mV (Figure 2A–C), indicates that even though features of the transmembrane domains are likely to set some voltage-depending gating properties, much of the varied voltage responses among BacNa_Vs arises from differences in neck domain properties. These gating effects are directly correlated with the ability of the neck to adopt a stable structure. Studies of chimeras containing the Na_VSp1 transmembrane domain and CTDs from other BacNa_Vs show that channels having the Na_VBh1 neck, which is disordered (Powl et al., 2010), have a voltage sensitivity that is equivalent to those lacking the entire CTD (Figures 1H and 2C), whereas those having a completely ordered CTD, Na_VAe1 (Shaya et al., 2014), require stronger depolarizations to open. Further, structural and functional studies show that increased neck disorder shifts the activation of both Na_VAe1 and Na_VSp1 to more hyperpolarized potentials (Figure 3 and (Shaya et al., 2014)). In this regard, our discovery of the ‘π-stack’ halogen-binding site (Figures 5D–E) provides a clear explanation for why Na_VAe1, among all other BacNa_Vs, can only be opened if the neck is destabilized.

Previous deletion studies of different BacNa_Vs reported varied effects, including a left shift in voltage-dependent activation for Na_VSulP (Irie et al., 2012) and no change for Na_VBh1 (Mio et al., 2010), and provided no clear view of why there were different outcomes in different channel contexts. Our data now point to a unified mechanism for BacNa_V gating in which the CTDs tune the ease of voltage-dependent opening of the channel based on the propensity of the neck region to adopt an ordered state. Due to its high polar residue content, the neck structure is metastable and contrasts with the stable hydrophobic core of the terminal coiled coil. This property is critical for the neck to undergo the order → disorder transition associated with channel opening (Figure 7A). Although CTD mutations can impact channel inactivation properties (Figures S1B–D and (Bagneris et al., 2013; Irie et al., 2010; Irie et al., 2012; Tsai et al., 2013)), the effects are diverse in different BacNa_Vs and likely reflect a complex inactivation process that also involves key contributions from elements of the pore domain (Pavlov et al., 2005; Payandeh et al., 2012; Shaya et al., 2014). Thus, the clear systemic effects we observe here regarding how neck structure affects activation leads us to propose a gating mechanism in which the order of the neck is directly linked to the ease of channel opening.

Figure 7. Function and structural conservation of VGIC superfamily CTD four-helix bundles, Related to Table S5.

A, BacNa_V gating is coupled to a Neck unfolding transition. Purple circles indicate Neck and Coiled-coil ions. B, Na_VAe1p (left) and TRPA1 (Paulsen et al., 2015) (right) cartoon diagrams CTD four-helix bundle is indicated. C, Na_VAe1p and TRPA1 coiled-coil superposition. TRPA1 shading indicates regions corresponding to the Na_VAe1p neck (light grey) and coiled-coil (dark grey). D, Comparison of Na_VAe1p (left) and TRPA1 (right) CTD cores. TRPA1 CTD buried hydrophilic residues are Gln1047 and Gln1061. E, Na_VAe1p and TRPA1 CTD superhelix radii comparison. Colors are as in ‘C’. See also Table S5.

Our studies revealed two ion binding sites that contribute to CTD functional properties and neck rigidity. The Na_VAe1 ‘π-stack’ motif is found in other halophile BacNa_Vs (Figure S7) and may serve as a point of ion-mediated modulation. The coiled-coil ion binding site comprises an E/Q/D/N/A-A-R motif at successive ‘a’-‘d’-‘a’ positions and is more prevalent among BacNa_Vs (Figure S7). This structure forms the lower boundary of loss of structure in neck disruptions (Figures 3A–B, and S6E) and provides a structural transition between the hydrophobic core of the coiled-coil and more open, hydrophilic core of the neck. The modular nature of the BacNa_V CTD and relatively simple four-helix architecture points towards means to rationally engineer channel responses by controlling the stability of the neck domain using either chimeric approaches or by exploiting stabilizing motifs such as the π-stack.

Some VGIC superfamily members, such as TRPs (Brauchi et al., 2006; Clapham and Miller, 2011; Grandl et al., 2008; Vriens et al., 2014) and K_2Ps (Bagriantsev et al., 2012; Lolicato et al., 2014; Maingret et al., 2000; Schneider et al., 2014) are gated by temperature. Whether ion channel temperature dependent responses originate from the action of a single domain (Bagriantsev et al., 2012; Brauchi et al., 2006; Grandl et al., 2008) or a more distributed property (Chowdhury et al., 2014; Clapham and Miller, 2011; Vriens et al., 2014) remains under intense investigation. The Na_VSp1 activation V_1/2 has a strong temperature dependence (~35 mV over a ~17°C range) that is eliminated by increasing neck disorder (Figure 2D–E). Further, Na_VSp1 CTD chimeras having a similar, Na_VSp1_PzCTD, or more right shifted, Na_VSp1_Ab1CTD, V_1/2 responses also show a strong temperature dependence, suggesting that the activation threshold position coincides with a more structured neck. These acute temperature responses contrast with the <5 mV response of the Shaker voltage-gated potassium channel over a similar temperature range (Chowdhury et al., 2014). Although distributed elements may influence BacNa_V temperature responses (DeCaen et al., 2014), particularly in Na_VBh1 (NaChBac), which has a disordered CTD (Powl et al., 2010), our data demonstrate that it is possible for a single domain, the BacNa_V CTD neck, to control the temperature dependent responses of a VGIC (Figure 2D–E). This result provides a definitive example of a defined temperature-sensing domain.

Structural studies of BacNa_Vs (Shaya et al., 2014), KCNQ channels (Howard et al., 2007; Wiener et al., 2008), and TRP channels (Paulsen et al., 2015; Yu et al., 2012) establish the commonality of parallel four-stranded coiled-coil domains in the VGIC superfamily. Comparison of the Na_VAe1p and TRPA1 CTDs reveals striking similarity well beyond their shared location (Figure 7B–C). Superposition shows a close correspondence in the coiled–coil (Figure 7C–D, Table S5) and highlights Na_VAe1p neck helix splaying (Figure 7E). The more uniform TRPA1 coiled-coil has two layers of buried hydrophilic residues: an N-terminal one that matches the position of the Na_VAe1p neck and a second one corresponding to the neck:coiled-coil junction (Figure 7C–D). Buried hydrophilic residues in coiled-coil cores carry a well-established energetic penalty with respect to quaternary architecture stability (Lupas and Gruber, 2005). Although TRPA1 ankyrin repeats have been implicated in temperature responses (Cordero-Morales et al., 2011; Jabba et al., 2014), it is striking that they form a cage around the CTD (Figure 7B). Hence, many of the reported ankyrin repeat domain effects may be due to modulation of CTD transitions. It is also notable that the voltage-sensor only channel Hv1 has a similar temperature sensitive membrane proximal coiled-coil domain bearing a buried polar residue in the region implicated in temperature responses (Fujiwara et al., 2012; Takeshita et al., 2014). Thus, the presence of buried polar residues in helical bundle domains in distantly related VGIC superfamily members together with the demonstration that such a metastable domain can play a crucial role in controlling channel function suggests that the mechanism we describe for the BacNa_V CTD may occur in many VGIC superfamily members.

Control of ion channel function by physical and chemical cues is essential for producing dynamic changes in excitability (Hille, 2001). While the coupling between the voltage sensor and pore domains has been studied extensively (Lu et al., 2002; Payandeh et al., 2012; Yifrach and MacKinnon, 2002), mechanisms for how cytoplasmic domains modulate the pore remain imperfectly understood. The four-helix bundle architecture of the BacNa_V CTD provides a simple means to control channel gating. The metastable, but ordered, neck structure directly couples to the energetics of pore opening (Figure 6). This effect does not depend on voltage-sensor movement and likely affects a late step of channel activation. These observations support the idea conformational changes in such structures can restrain and shape the energetics of channel pore opening (Shaya et al., 2014; Uysal et al., 2009). Consequently, helical bundle domains that bear buried polar residues proximal to a VGIC pore can provide a general mechanism for controlling the action of channels gated by different types of stimuli.

Experimental procedures

Construct design and cloning

Na_VSp1 (Silicibacter pomeroyi), Na_VAe1 (Alkalilimnicola ehrlichii), chimeras, and mutants were cloned into pIRES2-EGFP vector (Clontech, Mountain View, CA) (Shaya et al., 2014). Chimeric constructs were generated by PCR. Construct boundaries are found in Supplementary Materials. Mutants were made using QuikChange®(Stratagene). All constructs were verified by complete DNA sequencing.

Patch-clamp electrophysiology

Overexpression and patch-clamp recording from NavSp1, Na_VAe1, mutants, and chimeras were performed as described (Shaya et al., 2014). Voltage dependence was analyzed with the Boltzmann equation, y = 1/(1 + exp[(V V_1/2)/s]), where y is fractional activation, V is voltage, V_1/2 half-activation voltage and s is the inverse slope factor (mV). Details are found in Supplementary Materials.

Surface biotinylation assay

Surface expression was measured in HEK293 cells plated for each HA-tagged construct at 4°C. Samples were analyzed on SDS-PAGE and Western Blot detected. Details are found in Supplementary Materials.

Protein purification and crystallization

Na_VAe1p was expressed and purified as described previously (Shaya et al., 2014; Shaya et al., 2011) except that for the size exclusion chromatography step the running buffer included 150 mM NaCl, 50 mM CaCl₂, 0.3 mM β-dodecyl maltoside (DDM), 10 mM Na-HEPES, pH 8.0. Constructs for Na_VAe1p-3G and Na_VAe1p-7G were expressed in Escherichia coli (DE3) C43 (Miroux and Walker, 1996) using a previously described Na_VAe1p vector (Shaya et al., 2011). Details of expression, purification, and preparation of selenomethionine labeled Na_VAe1p-3G are found in Supplementary Materials.

Crystallization and structure determination

Na_VAe1p crystals were grown by hanging drop vapor diffusion. Na_VAe1p-3G, Na_VAe1p-3G SeMet, and Na_VAe1p-7G crystals were grown by microbatch under oil. Na_VAe1p, Na_VAe1p-NaBr complexes, and Na_VAe1p-3G SeMet diffraction data were collected at Advanced Photon Source Beamline 23ID-B, Argonne National Laboratory. Na_VAe1p-3G and Na_VAe1p-7G diffraction data were collected at Advanced Light Source Beamline 8.3.1, Lawrence Berkeley National Laboratory. Details for crystallization, data collection, structure determination, and model refinement can be found in the Supplementary Material.

Circular dichroism spectroscopy

CD spectra were recorded on an Aviv 215 spectrometer in a 1 mm pathlength quartz cell at 4 °C. BacNa_v pore domains were purified as above, exchanged into 10 mM sodium phosphate pH 7.4, 100 mM NaCl, 0.3 mM β-DDM, and concentrated to ~24–26 μM. Wavelength scans were recorded in triplicate from 190 to 320 nm with a 1 nm step size averaged over 10s. Thermal melts were as described (Shaya et al., 2011).

EPR labeling and sample preparation

Single cysteine mutants of Na_VAe1p, Na_VAe1p-3G, and Na_VSp1p were generated by quick-change mutagenesis and were purified as above using 1 mM TCEP in all buffers and labeled with MTSSL. CW-spectra were collected on a Bruker EMX at 10 mW power with a modulation amplitude of 1.6G. All spectra were normalized to the double integral. DEER experiments were carried out using a standard four-pulse protocol (Jeschke, 2002). Details can be found in the Supplementary Material.