Gating control of the cardiac sodium channel Nav1.5 by its β3-subunit involves distinct roles for a transmembrane glutamic acid and the extracellular domain

The auxiliary β3-subunit is an important functional regulator of the cardiac sodium channel Nav1.5, and some β3 mutations predispose individuals to cardiac arrhythmias. The β3-subunit uses its transmembrane α-helix and extracellular domain to bind to Nav1.5. Here, we investigated the role of an unusually located and highly conserved glutamic acid (Glu-176) within the β3 transmembrane region and its potential for functionally synergizing with the β3 extracellular domain (ECD). We substituted Glu-176 with lysine (E176K) in the WT β3-subunit and in a β3-subunit lacking the ECD. Patch-clamp experiments indicated that the E176K substitution does not affect the previously observed β3-dependent depolarizing shift of V½ of steady-state inactivation but does attenuate the accelerated recovery from inactivation conferred by the WT β3-subunit. Removal of the β3-ECD abrogated both the depolarizing shift of steady-state inactivation and the accelerated recovery, irrespective of the presence or absence of the Glu-176 residue. We found that steady-state inactivation and recovery from inactivation involve movements of the S4 helices within the DIII and DIV voltage sensors in response to membrane potential changes. Voltage-clamp fluorometry revealed that the E176K substitution alters DIII voltage sensor dynamics without affecting DIV. In contrast, removal of the ECD significantly altered the dynamics of both DIII and DIV. These results imply distinct roles for the β3-Glu-176 residue and the β3-ECD in regulating the conformational changes of the voltage sensors that determine channel inactivation and recovery from inactivation.

The auxiliary ␤3-subunit is an important functional regulator of the cardiac sodium channel Nav1.5, and some ␤3 mutations predispose individuals to cardiac arrhythmias. The ␤3-subunit uses its transmembrane ␣-helix and extracellular domain to bind to Nav1.5. Here, we investigated the role of an unusually located and highly conserved glutamic acid (Glu-176) within the ␤3 transmembrane region and its potential for functionally synergizing with the ␤3 extracellular domain (ECD). We substituted Glu-176 with lysine (E176K) in the WT ␤3-subunit and in a ␤3-subunit lacking the ECD. Patch-clamp experiments indicated that the E176K substitution does not affect the previously observed ␤3-dependent depolarizing shift of V1 ⁄2 of steady-state inactivation but does attenuate the accelerated recovery from inactivation conferred by the WT ␤3-subunit. Removal of the ␤3-ECD abrogated both the depolarizing shift of steady-state inactivation and the accelerated recovery, irrespective of the presence or absence of the Glu-176 residue. We found that steady-state inactivation and recovery from inactivation involve movements of the S4 helices within the DIII and DIV voltage sensors in response to membrane potential changes. Voltageclamp fluorometry revealed that the E176K substitution alters DIII voltage sensor dynamics without affecting DIV. In contrast, removal of the ECD significantly altered the dynamics of both DIII and DIV. These results imply distinct roles for the ␤3-Glu-176 residue and the ␤3-ECD in regulating the conformational changes of the voltage sensors that determine channel inactivation and recovery from inactivation.
The voltage-gated sodium (Nav) channel is an intrinsic plasma membrane protein complex that initiates the action potential in electrically excitable cells. The Nav channel is composed of a 250-kDa ␣-subunit, which consists of four homologous domains (DI-DIV) containing the voltage-sensing domains (VSDs) 5 and the ion-selective pore, in association with one or more auxiliary ␤-subunits that modify the gating behavior of the channel (1). In cardiac tissue, Nav1.5 is the predominant isoform. Mutations in Nav1.5 predispose to potentially fatal cardiac arrhythmias, such as the long QT and Brugada (BrS) syndromes (2). There is increasing evidence that the auxiliary ␤3-subunit is a major physiological regulator of the Nav1.5 ␣-subunit. For example, mice lacking ␤3 expression exhibit BrSlike cardiopathologies (3,4). Furthermore, several BrS patients have been identified with mutations in the ␤3-subunit (5)(6)(7)(8).
The ␤3-subunit contains an N-terminal extracellular domain (ECD), consisting of an Ig domain connected to a short, flexible, and probably disordered neck. This region is connected to a single transmembrane ␣-helix and a predominantly disordered intracellular C-terminal region (9, 10) (see Fig. 1). Recent structural evidence suggests that the DIII and DIV VSDs are implicated in the control of Nav channel inactivation (11,12). Voltage-clamp fluorometry (VCF) experiments strongly imply that in Nav1.5, the ␤3-subunit influences the gating behavior of both these VSDs (13,14). Consistent with this evidence, the ␤3-subunit accelerates Nav1.5 recovery from the inactivated state and produces a shift in the voltage, V1 ⁄ 2 , of steady-state inactivation (13,15). Although reports vary with respect to the direction of the voltage shifts, depending on the cell expression system used, they remain consistent with the ␤3-subunit altering the electrical field seen by one or more of the channel VSDs (16).
Mutagenesis and deletion experiments indicate that the ␤3-dependent shift in the V1 ⁄ 2 of inactivation requires the pres-ence of the ECD (15). However, the role of the ␤3 transmembrane ␣-helix has been less well-studied. We have previously noted the presence of a highly conserved glutamic acid residue within the transmembrane region at position 176 (16). The location is intriguing because glutamic acids are rarely located in such hydrophobic regions (17). Here we mutate the Glu-176 residue to lysine in both the WT ␤3-subunit and the ␤3-subunit lacking the ECD. We examine the effects of these mutations on gating behavior and voltage-dependent movements of Nav1.5 DIII and DIV VSDs, using whole-cell patch-clamp electrophysiology and VCF. Our results suggest that the ␤3-subunit transmembrane ␣-helix lies close to the DIII VSD, where the Glu-176 residue plays an important role in facilitating recovery from inactivation. We further show that the ␤3-subunit extracellular region extends its influence both to the DIII and DIV VSDs. These data provide new structural and functional insights into how the ␤3-subunit can modulate both steady-state inactivation and recovery from inactivation.

Structural insights into the Glu-176 residue and its location within the transmembrane domain
The amino acid residues referred to in the text are numbered from the start codon of the cDNA. Thus, the endoplasmic reticulum targeting signal is included in the numbering convention. Secondary structure predictions identified the likely position and extent of the transmembrane region as a conserved, hydrophobic, ␣-helical region extending from residues 156 to 180. The transmembrane region of single-pass, type 1 membrane proteins such as the ␤3-subunit are often enriched for residues such as tyrosine and cysteine on the membrane side of the cytoplasmic interface, followed on the cytoplasmic side by a cluster of basic residues (18). All these features are present in the ␤3 transmembrane region at positions 179 -183 (Fig. 1A). Hence, the ␤3 transmembrane ␣-helix is likely to emerge into the cytosol at or close to residues Arg-182 and Lys-183. If so, the Glu-176 residue will be located almost two ␣-helical turns, or 9 -10 Å, inside the membrane. Interestingly, the Glu-176 residuewhich is fully conserved between all known ␤3 sequences-lies on a face of the ␣-helix that is also fully conserved (Fig. 1, A and  B). This suggests the presence of an extended binding site along the transmembrane ␣-helix, in which the Glu-176 residue is likely to play an important role.
To investigate the function(s) of the Glu-176 residue, we used site-directed mutagenesis to replace it with lysine in the full-length ␤3 sequence. The lysine side chain has similar dimensions to the glutamic acid side chain but in aqueous media carries a positive charge. In a hydrophobic environment, the pK a of a lysine side chain is between 6 and 6.5 (19,20). Hence, even if buried within the membrane, a significant minority of the lysine should remain positively charged and thus likely to be electrostatically distinct from the WT glutamic acid. To investigate the potential for functional synergy between the Glu-176 residue and the extracellular domain, we also introduced the E176K mutant into a ␤3 deletion mutant lacking the extracellular domain (15). All ␤3 constructs contained a C-terminal EGFP tag to aid the identification of trans-fected cells. A diagrammatic summary of the mutational constructs used in this work and their naming convention is shown in Fig. 1C.

The Glu-176 residue is not required for ␤3-subunit self-oligomerization
The isolated ␤3-subunit Ig domain can form trimers in solution and the full-length ␤3-subunit forms trimers in the plasma membrane when expressed in HEK293 cells in the absence of the Nav channel ␣-subunit (21). We have previously suggested that the ␤3-subunit trimer could be stabilized by hydrogen bonding between the membrane-embedded and protonated Glu-176 side chains on adjacent transmembrane domains (16). To test this hypothesis, cross-linking experiments with the membrane-impermeant bis-sulfosuccinimidyl suberate (BS3) were carried out on HEK293F cells singly transfected with the full-length WT ␤3-subunit and the different ␤3-subunit mutations ( Fig. 2A).
In the absence of the cross-linker BS3, WT-␤3-EGFP and E176K-EGFP migrated as monomers on SDS-PAGE ( Fig. 2A). The presence of closely spaced bands is due to multiple glycosylation states (22). Similarly, under these conditions the ␤3-⌬ECD-EGFP and the ␤3-⌬ECD-E176K-EGFP proteins were observed as monomers at ϳ35-38 kDa, In the presence of BS3, extra bands were observed for the WT-␤3-EGFP and E176K-EGFP ␤3-subunits at ϳ130 -140 and 180 -200 kDa, indicating the formation of dimers and trimers. There was no obvious reduction in the extent of cross-linking between WT and E176K mutant, suggesting that the mutation does not prevent the ability to form dimers and trimers. However, oligomerization induced by cross-linking was lost with the removal of the ECD, whether or not the resulting ⌬ECD construct contained the E176K mutation in its transmembrane region. Hence, oligomerization depends more on the presence of an intact ECD than on the transmembrane region. We found no evidence that the homophilic interactions between individual ␤3-subunits specifically required the presence of a glutamic acid at position 176.

Binding of the ␤3-subunit to the Nav1.5 ␣-subunit persists with the E176K mutation
Binding of the different ␤3-subunit variants to the Nav1.5 ␣-subunit was examined by co-immunoprecipitation experiments. Here, a Nav1.5 ␣-subunit with a C-terminal HA tag (21) was co-transfected with the WT or mutant ␤3-subunits. This resulted in the co-immunoprecipitation of all co-expressed ␤3-subunits, including the ␤3-⌬ECD-E176K-EGFP mutant that both lacked the extracellular Ig domain and contained the E176K change (Fig. 2B). This suggests that neither the extracellular Ig domain nor the Glu-176 residue alone are essential for the binding of the ␤3-subunit to the Nav1.5 ␣-subunit. Interestingly, it was noticeable that the intensity of the immunoprecipitated ␤3-E176K EGFP mutant in the bound fraction was somewhat weaker compared with WT-␤3 EGFP, whereas the immunoprecipitated ␤3-⌬ECD (with or without the E176K mutant) was noticeably more intense. Hence, the Glu-176 residue and the ECD may influence the overall stability of the Nav1.5-␤3 complex with Nav1.5 in different ways (see "Discussion"). Stable Nav1.5-HEK293F cells were co-transfected with an empty EGFP vector (Nav1.5 with EGFP) or EGFP-tagged WT or mutant ␤3-subunits (i.e. Nav1.5 with WT-␤3-EGFP; Nav1.5 with ␤3-E176K-EGFP; Nav1.5 with ␤3-⌬ECD-EGFP; and Nav1.5 with ␤3-⌬ECD-E176K-EGFP). Representative traces of whole-cell sodium currents (I Na ) in response to an activation protocol are shown in Fig. 3A. None of the ␤3-subunit variants resulted in significant changes in peak I Na , voltage, V1 ⁄ 2 of maximal activation or slope factor, and k (p Ͼ 0.05; Fig. 3, B-D, and Table 1).
Typical inactivation traces are shown in Fig. 4, and the parameters are summarized in Table 1. Expression of WT-␤3-EGFP resulted in a ϳ7-mV depolarizing shift of Nav1.5 steadystate inactivation, agreeing with previous findings (13) expressing that ␤3-E176K-EGFP did not alter this effect. However, the loss of the extracellular Ig domain completely abolished this shift, independent of the presence of the Glu-176 residue. The slope factor, k, exhibited a small but significant decrease in the presence of WT-␤3-EGFP, suggesting a slight increase in voltage dependence of inactivation.

Both loss of the ECD and the E176K mutation abrogate the ␤3-mediated acceleration of recovery from Na ؉ current inactivation
It has previously been shown that the ␤3-subunit accelerates Na ϩ channel recovery from inactivation (13,23). We sought to determine whether either the extracellular Ig domain or the Glu-176 residue influenced this mechanism. Representative whole-cell Na ϩ currents were elicited by a double-pulse protocol, whose pulses were separated by progressively incremental time intervals, ⌬t, are shown in Fig. 5. These respective P1 and P2 pulses yielded peak currents I P1 and I P2 . The degrees of recovery following different time intervals, ⌬t, were compared by obtaining the ratio of I P2 /I P1. Plots of I P2 /I P1 as a function of time are fitted to a double exponential, comprising a fast and slow component, as described under "Experimental procedures." Both the fast and slow components of Nav1.5 recovery were accelerated by the co-expression of WT-␤3-EGFP compared with those obtained with the co-expression of Nav1.5 with EGFP alone (p Ͻ 0.005; Fig. 5 and Table 2). The ␤3-E176K mutation significantly attenuated both of these effects but did not completely abolish it. The loss of the extracellular Ig domain also resulted in a slowing of recovery from inactivation compared with WT-␤3-EGFP. That was more pronounced in the fast component. Hence, the kinetic and steady-state data together indicate overlapping but distinct roles for the Glu-176 residue and the extracellular Ig domain of ␤3 in regulating recovery from inactivation.

Distinct roles for the Glu-176 residue and the ECD in modulating DIII and DIV VSD movement
The Nav channel VSDs are comprised of the first four transmembrane ␣-helices (S1-S4) within each of the four internally homologous Nav channel domains (DI-DIV) (Fig. S1). Changes in membrane potential initiate movement of the charged S4 helix within each VSD. This enables channel opening (for the case of DI-DIII VSDs) and subsequent channel inactivation (for the case of DIV VSD) (1). The regulatory effect of the ␤3-subunit on recovery from inactivation arises at least in part by its modulation of the voltage responses of the DIII and DIV VSDs (13,14). Further investigations into the effect of the Glu-176 residue upon these two VDSs utilized VCF in cut-open Figure 1. Sequence analysis of ␤3-structures. A, secondary structure prediction of the ␤3-subunit transmembrane region and its sequence alignment among a wide range of vertebrate species. Residue numbers refer to the human sequence. The predicted transmembrane region is colored orange, and the Glu-176 residue is highlighted. Residues fully conserved between species are indicated with asterisks. B, space-filling model of the ␤3-subunit with the fully conserved residues of the transmembrane region shown in magenta. Note that the helix face containing the Glu-176 residue is fully conserved. Analysis and modeling were as described under "Experimental procedures." C, cartoon summary of the WT and mutant ␤3-subunit constructs used in this work and referred to in the text. ␤3 structures influence Nav1.5 gating oocytes. Two separate constructs (M1296C and S1618C) were generated in the Nav1.5 ␣-subunit by site-directed mutagenesis. The M1296C introduced a cysteine into the extracellular loop connecting helices S3 and S4 of the DIII VSD; the S1618C mutation introduced a cysteine into the extracellular loop connecting helices S3 and S4 of the DIV VSD (13, 24) (Fig. S1).
These cysteines were selectively labeled with the fluorophore MTS-TAMRA. Changes in fluorescence intensity (⌬F) of the TAMRA adjunct were used to track conformation changes of the S4 helix in the DIII and DIV VSDs in response to changes in membrane potential (13). Previous experiments established that cysteine introduction and labeling of the M1296C and  Table 1 for individual groups), and statistical significance was tested with one-way ANOVA. All parameters (peak I Na , V1 ⁄2 , and k) showed no statistically significant variation (p Ͼ 0.2). See Table 1 for individual values. Table 1 Nav1.5 activation and steady-state inactivation and recovery from inactivation parameters with and without the ␤3 WT and mutant subunits Activation (G/G max ) and inactivation data are fit to the Boltzmann function, and V1 ⁄ 2 and k were derived from this. Peak I Na is the mean of the absolute maximum I Na elicited by each cell during the activation protocol. All data are means Ϯ S.D. (n Ն 6, indicated in the table), compared using one-way ANOVA (p Ͼ 0.2 for all activation parameters, and p Ͻ 0.01 for inactivation parameters). Parameters that were determined to be statistically significant were subjected to a Sidak's multiple comparison post hoc test (all conditions were compared against Nav1.5 ϩ EGFP and Nav1.5 ϩ ␤3-EGFP). ␤3 structures influence Nav1.5 gating S1618C Nav1.5 constructs caused only small changes in channel gating properties compared with WT Nav1.5 (24). Moreover, co-expression of the ␤3-subunit with MTS-TAMRA labeled M1296C and S1618C Nav1.5 constructs caused similar gating shifts as with the WT Nav1.5 channel (13). On this basis, we conclude that the interactions between Nav1.5 ␣-subunit and the ␤3-subunit are not fundamentally compromised by the VCF mutations or the MTS-TAMRA labeling.
␤3 structures influence Nav1.5 gating tant in mediating the effect of the ␤3-subunit on DIII VSD modulation. In contrast, the ␤3-⌬ECD-EGFP mutant and the ␤3-⌬ECD-E176K-EGFP double mutant activated the DIII VSD at more hyperpolarized potentials; this was such a drastic effect that it produced a significantly hyperpolarized shift of V1 ⁄ 2 even compared with Nav1.5 expressed in the absence of the ␤3-subunit. However, there does not appear to be an additive effect in the double mutation ( Fig. 6 and Table 3). The slope factor (k), which is inversely proportional to charge movement across the membrane (see "Discussion"), was almost doubled by the ␤3-E176K-EGFP mutant, yet the ␤3-⌬ECD-EGFP and the ␤3-⌬ECD-E176K-EGFP mutants, as well as the WT-␤3-EGFP subunit, had little effect on this parameter compared with Nav1.5 expressed alone ( Fig. 6 and Table 3).
For the case of the DIV VSD, the WT-␤3-EGFP and mutants had little effect on V1 ⁄ 2 for the S4 helix movement relative to Nav1.5 ␣-subunit alone. However, WT-␤3-EGFP caused the initial activation of the DIV-VSD to occur at a more positive potential than in the absence of ␤3 and showed a greater voltage dependence of activation, with a reduced slope factor for the S4 helix movement in DIV VSD. These effects of ␤3 were not altered by the presence of the E176K mutation. The ␤3-⌬ECD-EGFP single mutation activated the DIV-VSD at more negative potentials compared with WT-␤3-EGFP. In addition, there was a significantly decreased voltage dependence of activation with increased slope k values compared with WT-␤3-EGFP and E176K-EGFP, whereas the ␤3-⌬ECD-E176K-EGFP double mutant produced a still greater increase in k ( Fig. 6 and Table 3).
Taken together, these results suggest that the ␤3-subunit Glu-176 residue and the ECD both influence movement of the DIII VSD, albeit through distinct mechanisms. In addition, the ECD is also important in mediating the effects of the ␤3-subunit on the voltage dependence of the DIV VSD movement (Fig. 6 and Table  3), translating to the depolarizing shift of V1 ⁄ 2 inactivation exhibited with co-expression of full-length WT ␤3-subunit.

Discussion
Previous work has shown that distinct structural regions and domains within the ␤3-subunit modify discrete aspects of Nav channel gating behavior (10,15,25). Here, we explore the functional significance of additional structural features within the ␤3-subunit protein: in particular, the Glu-176 residue within the transmembrane ␣-helix.
Glutamic acid is significantly under-represented within ␣-helical transmembrane sequences. This relative rarity probably reflects the thermodynamic cost of burying a potentially negatively charged side chain in a hydrophobic environment (18). In principle, such an unfavorable interaction could be compensated by allowing water molecules to surround the side chain or by an increase in the pK a of the glutamic acid side chain so that it remains largely protonated and uncharged when inside the membrane. However, both possibilities could have important structural consequences (17). Given these constraints, the striking evolutionary conservation of the Glu-176 residue strongly argues for its functional significance. Interestingly, the pattern of maximum sequence conservation throughout the entire transmembrane ␣-helix shows an approximate 4:3 periodicity. This will generate a helix where the maximally conserved residues, including Glu-176, all lie on the same face (Fig. 1, B and C). We therefore suggest that the Glu-176 residue is part of a larger feature, most likely a binding or interaction site.
The Nav channel ␤-subunits are members of the immunoglobulin superfamily and are related to members of the celladhesion N-CAM family. Indeed, Nav ␤-subunits can function as cell-adhesion molecules independently of the Nav ␣-subunits (26). Nav ␤-subunits can also adopt multiple oligomeric states in the absence of Nav ␣-subunits. For example, the ␤2 and ␤4-subunits can assemble as both cis and trans oligomers to promote trans cell adhesion (27). Similarly, the full-length ␤3-subunit can form cis-interacting dimers and trimers when expressed in cells without the Nav ␣-subunit (21). We have previously suggested that the membrane-embedded, protonated form of the Glu-176 residue could help stabilize these ␤3-subunit dimer and trimers by intersubunit hydrogen bonding (16,21). However, our new data indicate that oligomerization of the full-length ␤3-subunit does not depend on the presence of the Glu-176 residue but depends on the presence of an intact ECD (Fig. 2A). This is consistent with structural data showing the purified ECD immunoglobulin domain can form trimers in vitro (21) and implies that the Glu-176 residue has been conserved for other purposes.
The location of the physiologically relevant ␤3-binding site on the Nav1.5 ␣-subunit is currently uncertain. Recent cryo-EM structures for the eel Nav channel, human Nav1.4, and human

recovery from inactivation parameters with and without the ␤3 WT and mutant subunits
Recovery from inactivation data is fit to a double exponential function. All data are means Ϯ S.D. (n Ն 7, indicated in the table). Statistically significant results were determined using one-way ANOVA (all parameters p Ն 0.0006) and compared with a Sidak's multiple comparison post hoc test (all conditions were compared against Nav1.5 ϩ EGFP and Nav1.5 ϩ ␤3-EGFP). ␤3 structures influence Nav1.5 gating Nav1.7 have identified the binding site for the ␤1-subunit transmembrane region between the S1 and S2 helices of the DIII VSD (11,28,29). The transmembrane regions of the ␤1 and the ␤3-subunits share significant sequence similarity, and this includes a glutamic acid in the ␤1 sequence at the equivalent position to Glu-176 in ␤3 (9). It is therefore tempting to assume that the ␤1 and ␤3 transmembrane regions may bind to the same site, and previous VCF experiments do indeed indicate that the ␤3-subunit transmembrane region binds within the DIII VSD (13). However, several independent lines of evidence suggest that this hypothesis may be too simplistic. For example, in contrast to the results presented here, the equivalent Glu to Lys mutation in the transmembrane region ␤1 induced a significant shift in V1 ⁄ 2 for steady-state inacti-vation but had no effect on the recovery from inactivation (30).

Recovery from inactivation
Then VCF experiments showed that unlike ␤3, the ␤1-subunit did not affect V1 ⁄ 2 or k for fluorescence changes in DIII for Nav1.5 (13). Furthermore, an introduced tryptophan in the ␤3-ECD neck connecting the Ig domain to the transmembrane region was able to quench the fluorescence signal from MTS-TAMRA selectively tagged at position 1296 of Nav1.5. This places the ␤3 neck region no more than 15 Å from residue 1296 of the Nav1.5 DIII, S3-S4 extracellular loop (13), a distance that is too short to be compatible with the known location of the ␤1-subunit transmembrane region on DIII between S1 and S2 (11,12). Crucially, the same fluorescence quenching experiments, using an equivalent tryptophan mutation on the ␤1-subunit failed to detect quenching from  Table 3) compared by one-way ANOVA (all parameters p Յ 0.0005) and followed by a Sidak's multiple comparison post hoc test (all conditions were compared against Nav1.5 ϩ EGFP and Nav1.5 ϩ ␤3-EGFP; see Table 3 Table 3 has the full numerical values and statistical comparisons. ␤3 structures influence Nav1.5 gating Nav1.5 tagged with MTS-TAMRA at position 1296 (13). On this basis, we propose that the ␤3 transmembrane ␣-helix is more likely to lie between the Nav1.5 DIII S3 and S4 helices. This could allow the ␤3 ECD to interact with the DI-S5 and DIV-S5 extracellular loops but in a different orientation compared with ␤1 ECD. As a result, the full-length ␤3-subunit could potentially influence both DIII and DIV in a way that is distinct from ␤1.
As judged by immunoprecipitation, the ␤3-E176K-EGFP mutation interacted with the Nav1.5 ␣-subunit, although the amount of E176K ␤3 co-precipitated with Nav1.5 was somewhat reduced compared with WT ␤3 (Fig. 2B). This could reflect a somewhat lower affinity of the E176K ␤3 mutant for the Nav1.5 ␣-subunit, leading to lower retention of the E176K ␤3 mutant during the immunoprecipitation and extensive washing steps. However, both the E176K mutant and the WT ␤3-subunits induced a similar depolarizing shift in the V1 ⁄ 2 of Nav channel inactivation relative to that shown by the Nav1.5 ␣-subunit alone ( Fig. 4 and Table 1). Therefore, the E176K mutant was still capable of functionally interacting with the Nav1.5 ␣-subunit in vivo.
To investigate the roles of the ␤3-subunit Glu-176 residue and its ECD, we used both patch-clamp electrophysiology and VCF. These two approaches provide complementary but different insights into Nav channel behavior. The electrophysiology reflects a culmination of all the steps and conformational changes that must occur before the channel can activate or inactivate. On the other hand, VCF experiments provide a more specific insight into voltage-dependent movements that occur in the S4 helices of individual VSDs, which subsequently lead to activation and/or inactivation (31).
The VCF experiments showed that in the absence of the ␤3-subunit, the V1 ⁄ 2 for the movement of the DIII S4 helix was nearly 50 mV more hyperpolarized than the equivalent value for the DIV S4 helix. Hence, for Nav1.5 alone, a depolarizing signal will initiate S4 helix movement considerably sooner in DIII than in DIV, and there will necessarily be a delay between the initiation of activation and subsequent inactivation pathways. Strikingly, however, the presence of full-length ␤3-subunit considerably reduced the difference between the V1 ⁄ 2 values for the DIII and DIV S4 helix movements (Fig. 6B and Table 3). By reducing this difference, the ␤3-subunit ensured that the DIII and DIV S4 helix movements occurred over a narrower voltage range and thus generated a more coordinated depolar-ization response from these two VSDs. Almost all the ␤3-induced relative changes in the V1 ⁄ 2 values occurred with the TAMRA-tagged C1296 Nav1.5 construct, suggesting that the DIII VSD is likely to interact directly with the ␤3-subunit.
It is striking that, compared with the WT-␤3-EGFP, the fulllength E176K mutant abrogated the effect of ␤3 on the voltage sensitivity of fluorescence changes for DIII VSD, while significantly increasing the slope factor k ( Fig. 6B and Table 3). The slope factor, k, is inversely related to the amount of charge that can move across the membrane during a transition (32). Hence, the E176K mutant dramatically reduced this value, suggesting that without the ␤3-Glu-176 residue, the movement of the S4 helix in DIII during depolarization was compromised or at least altered relative to WT ␤3. A role for the Glu-176 residue in facilitating DIII S4 helix movement could also explain the effect of the ␤3-E176K-EGFP mutant on recovery from inactivation ( Fig. 5 and Table 2). The DIII VSD is known to mediate recovery from inactivation via its interaction with the DIII-DIV linker region (14). By correctly positioning the DIII S4 helix in the inactivated state, the Glu-176 residue could in effect catalyze the ␤3-dependent accelerated recovery from inactivation as the membrane potential is repolarized.
Removal of the ␤3-subunit ECD (both with and without the E176K mutant) fully abrogated the effects of WT ␤3-subunit on the V1 ⁄ 2 for voltage-dependent fluorescence changes in DIII VSD (Fig. 6, A and B, and Table 3). This suggests that the ECD may play a role in correctly positioning the ␤3-subunit transmembrane ␣-helix. Interestingly, the co-immunoprecipitation experiments indicated that ␤3 constructs lacking the ECD bound to Nav1.5 more intensely compared with WT-␤3-EGFP (Fig. 2B). It is therefore possible that the ECD may restrict the binding of the ␤3-subunit to one or a small number of physiologically appropriate sites on the Nav1.5 ␣-subunit. This could be achieved through the ECD Ig region interacting with specific extracellular loop regions on the ␣-subunit. Removal of the ECD could compromise this discrimination and enable the hydrophobic transmembrane region more freedom to bind additional sites on the ␣-subunit or to the same DIII region but with multiple orientations or conformational states, so that the Glu-176 residue is incorrectly positioned. In this case, the DIII S4 helix movement would no longer be so sensitive to the presence of the Glu-176 residue.
The DIV VSD plays an important role in steady-state inactivation (33). The WT-␤3-EGFP subunit reduced the value of k Table 3 VCF data for DIII and DIV in Nav1. 5

with and without the ␤3 WT and mutant subunits
The V1 ⁄ 2 and k values for DIII and DIV are derived from the Boltzmann fit of the normalized change in fluorescence as a function of voltage. All data are means Ϯ S.D. (n Ն 3 indicated in the table). Statistically significant results were determined using one-way ANOVA (all parameters p Յ 0.0005) followed by Sidak's multiple comparison post hoc test (all conditions were compared against Nav1.5 ϩ EGFP and Nav1.5 ϩ ␤3-EGFP). ␤3 structures influence Nav1.5 gating for voltage-dependent S4 helix movements in DIV VSD (Fig. 6, A and B, and Table 3). This change was abolished for ␤3-subunits lacking the ECD (with or without the E176K mutation).

DIII F-V DIV F-V
Removal of the ECD also abolished the ␤3-induced depolarizing shift in steady-state inactivation gating ( Fig. 4 and Table 1). Hence, the ␤3-ECD is likely to influence DIV, despite its location on the DIII VSD. One possibility is that the ␤3 ECD does this indirectly by modifying the conformation of the DI S5-S6 extracellular loop, adjacent to the DIV VSD. Alternatively, if the ␤3 transmembrane region binds to DIII VSD, it is likely that the ECD will interact with the DIV S5-S6 extracellular loop (Fig.  S1), which could provide a more direct coupling between ECD and DIV. Furthermore, the magnitude of voltage-dependent fluorescence changes in DIII and DIV VSDs recorded with both ␤3-⌬ECD-EGFP and ␤3-⌬ECD-E176K-EGFP were less pronounced compared with Nav1.5 alone or with full-length ␤3 constructs (Fig. 6A). Because the fluorescence signal is very sensitive to changes in the immediate environment of the fluorophore (34), it is possible that removal of the ECD subtly altered the trajectory of the DIII and DIV S4 helix movements compared with Nav1.5 alone or the full-length ␤3 constructs. In summary, our data provide new and quantitative insights into the unique role of the ␤3-subunit in the regulation of Nav1.5 gating. In this study, our emphasis has been to clarify the importance of distinctive and unusually positioned residues using the simplest and most tractable Nav1.5 ␣and ␤3-subunits alone. In the future, building on this current work, it will be interesting to investigate these mutations in a more complex physiological system such as a cardiomyocyte, using, for example, targeted gene replacement methods.

Secondary structure predictions and modeling
Secondary structure prediction of the ␤3-subunit transmembrane region was carried out using Phyre2 (35). The alignment analysis of the ␤3 transmembrane region among different species was carried out using Clustal Omega program (36). A structural model of the ␤3-subunit was generated using the I-TASSER server (37).

cDNA constructs, cell culture, mutagenesis, and cell transfections
Plasmids pcDNA3-Nav1.5-hemagglutinin (HA), pcDNA3 WT-␤3-EGFP, and pcDNA3 ␤3-⌬ECD-EGFP have been previously described (10,15,38). Site-directed mutagenesis of WT-␤3-EGFP and ␤3-⌬ECD-EGFP to introduce the E176K mutation was performed using the QuikChange Lightning kit (Agilent) with primers (designed complementary to the ␤3 sequence, except for the appropriate base change for glutamic acid to lysine (E176K) in the middle of the primer) from Sigma. Colonies were picked and grown in culture medium overnight, plasmids were then extracted using the Genelute plasmid midiprep kit (Sigma), and successful mutation was identified by Sanger sequencing (DNA sequencing facility, University of Cambridge). Separate Nav1.5 constructs containing the M1296C and S1618C were generated as previously described (24).

Co-immunoprecipitation
48 h post-transfection, the cells were washed three times in cold PBS and then lysed in 1 ml of lysis buffer (50 mM Tris, 150 mM NaCl, 1% Triton X-100 (v/v)) supplemented with protease inhibitor mixture (Roche, Sigma-Aldrich)). Lysates were vortexed, stored on ice for 30 min, and centrifuged at 10,000 ϫ g for 10 min at 4°C, and the pellet (cell debris) fraction was discarded. Lysates were incubated in either mouse anti-GFP tag or mouse anti-HA with end-over-end rotation at 4°C overnight, followed by the addition of protein G-agarose for 4 h. The samples were spun at 2,000 ϫ g at 4°C for 5 min. Pellet (bound) fractions were washed four times in 1 ml of lysis buffer, and both these and the supernatants (unbound) were incubated in 4ϫ NuPage loading buffer supplemented DTT at 85°C for 10 min. The bound and unbound fractions were separated on SDS-PAGE and transferred to nitrocellulose membranes, and Western blotting was carried out with rabbit polyclonal antibodies anti-GFP (GeneTex, Insight Biotech) and anti-HA (Santa Cruz, Insight Biotech) to detect EGFP as well as the ␤3-GFP WT and mutants and Nav1.5-HA, respectively.

BS3 cross-linking experiments
Transfected cells were washed three times in Hanks' balanced salt solution and then lysed in Hanks' balanced salt solution lysis buffer (1% Triton X-100, 0.2% SDS, 0.5% sodium deoxycholate) supplemented with protease inhibitors (Roche cOmplete Protease Inhibitor mixture). Each lysate was separated into two equal fractions, one of which was incubated with BS3 (5 mM) for 1 h at 4°C. The reaction was quenched with glycine (74 mM), and the lysate was clarified at 10,000 ϫ g for 10 min at 4°C. The lysates were then subjected to SDS-PAGE and transferred to nitrocellulose membranes, and Western blotting was carried out with rabbit polyclonal antibodies anti-GFP (GeneTex, Insight Biotech) and anti-myc (Santa Cruz, Insight Biotech) to detect ␤3-GFP and ␤3-myc, respectively.

Whole-cell patch clamp
Na ϩ currents (I Na ) were recorded from HEK293F cells stably expressing Nav1.5 and transiently transfected with 1 g of the ␤3 constructs or a vector containing only EGFP. Successfully transfected cells were identified by EGFP fluorescence on an Olympus IX71 inverted microscope. Experiments were carried out at room temperature (22-23°C) in the whole-cell configuration with an Axopatch 200B amplifier (Axon Instruments), a ␤3 structures influence Nav1.5 gating Digidata 1322A digitizer (Axon Instruments), and the Strathclyde Electrophysiology Software Package (WinWCP, Department of Physiology and Pharmacology, University of Strathclyde). The extracellular bath solution contained 60 mM NaCl, 2 mM KCl, 1.5 mM CaCl 2 , 10 mM glucose, 1 mM MgCl 2 , 90 mM CsCl 2 , 10 mM HEPES, pH 7.39 Ϯ 0.02 with NaOH. 1.5-2.5 M⍀ patch pipettes were produced from borosilicate glass capillaries (Harvard Apparatus Ltd.) using a horizontal puller (P-87; Sutter Instruments) and filled with intracellular saline comprised of 35 mM NaCl, 105 mM CsF, 10 mM EGTA, 10 mM HEPES, pH 7.39 Ϯ 0.02 with CsOH. Signals were low-pass Bessel filtered at a frequency of 5 kHz and sampled at 125 kHz. Series resistance compensation was performed to 75-80%, and leak currents were subtracted using a P/4 protocol. The liquid junction potential (2 mV) was not corrected for. Data from cells with a current amplitude larger than 8 nA or with a clear loss of voltage control as demonstrated by poor I/V relationships were removed.

Voltage protocols
All voltage protocols used a holding voltage of Ϫ120 mV. The steady-state inactivation and activation protocol consisted of a 100-ms depolarizing pulse ranging from Ϫ140 mV to ϩ35 mV, in 5-mV increments, followed by a fixed Ϫ40 mV depolarizing pulse of 50-ms duration. Currents elicited from the first pulse constitute activation data and those from the second depolarizing pulse provide inactivation data.
Current traces were normalized against the whole-cell capacitance (C m ), and the I/V relationship was plotted from the peak current at each test voltage. The values of Na ϩ conductance (G Na ), for families of traces at each test voltage, were determined from the equation, where I Na is the Na ϩ current and E Na is the Na ϩ reversal potential. Peak G Na was plotted as a function of voltage to produce activation curves. I Na was normalized to the maximum elicited current and plotted against the conditioning voltage to yield inactivation curves. Both curves were fitted to the following Boltzmann function, G/G max ϭ 1/͑1 ϩ exp͑͑V Ϫ V1 ⁄2 ͒/k͒͒ (Eq. 2) where G/G max is the normalized conductance or current, V1 ⁄ 2 is the voltage of half-maximal activation or inactivation, k is the slope factor, and V is the test voltage or conditioning voltage.
Recovery from inactivation was examined using a double-pulse P1 and P2 protocol that delivered two identical depolarizing pulses to Ϫ40 mV of 50-ms duration. The time interval between P1 and P2 was initially incremented by 1 ms up to 6 ms, followed by 2-ms increments to 20 ms, then 5-ms increments to 60 ms, followed by 10-ms increments to 120 ms, and finally 20-ms increments to 200 ms to ensure enough time was allowed for full recovery and to allow adequate capture of the fast components. Peak currents from P2 were normalized to those obtained in response to the conditioning P1 step and plotted against the time intervals. These plots were fitted with a double exponential function as follows, where t is the time, and is the time constant of recovery from inactivation.