Secondary structure of the human cardiac Na+ channel C terminus: evidence for a role of helical structures in modulation of channel inactivation.

Little is known about the structure of the C terminus of the human cardiac voltage-gated sodium channel alpha subunit (SCN5A), but disease-linked mutations within this 244-amino acid intracellular region of the channel have marked effects on channel inactivation. Here we report a structural analysis of the C-terminal tail of the cardiac Na(+) channel that sheds new light on mechanisms that control its inactivation gating. Homology modeling of the SCN5A C terminus predicts predominant alpha-helical structure (six helices) in the proximal half of this intracellular tail but little structure in the distal half. Circular dichroism of isolated and purified C terminus supports this prediction. Whole cell and single channel patch clamp recordings of wild type and mutant alpha subunits co-expressed with the hbeta(1) subunit in HEK 293 cells indicate that truncation of the distal, nonstructured, C terminus (L1921stop mutant) reduces current density but does not affect channel gating (n = 6). In contrast, truncation of the sixth helix containing a concentration of positively charged residues along with the distal C terminus (S1885stop mutant) also reduces current density but, in addition, has profound and selective effects on inactivation (no effect on activation). Channel availability is shifted (-11 +/- 0.6 mV), and there is a 10-fold increase in the percentage of channels that burst (fail to inactivate) during prolonged depolarization (0.025% S1885stop (n = 7) versus 0.0028% wild type (n = 9), p < 0.005). These results suggest that the charged structured region of the SCN5A C terminus plays a major role in channel inactivation, stabilizing the inactivated state.

Little is known about the structure of the C terminus of the human cardiac voltage-gated sodium channel ␣ subunit (SCN5A), but disease-linked mutations within this 244-amino acid intracellular region of the channel have marked effects on channel inactivation. Here we report a structural analysis of the C-terminal tail of the cardiac Na ؉ channel that sheds new light on mechanisms that control its inactivation gating. Homology modeling of the SCN5A C terminus predicts predominant ␣-helical structure (six helices) in the proximal half of this intracellular tail but little structure in the distal half. Circular dichroism of isolated and purified C terminus supports this prediction. Whole cell and single channel patch clamp recordings of wild type and mutant ␣ subunits co-expressed with the h␤ 1 subunit in HEK 293 cells indicate that truncation of the distal, nonstructured, C terminus (L1921stop mutant) reduces current density but does not affect channel gating (n ‫؍‬ 6). In contrast, truncation of the sixth helix containing a concentration of positively charged residues along with the distal C terminus (S1885stop mutant) also reduces current density but, in addition, has profound and selective effects on inactivation (no effect on activation). Channel availability is shifted (؊11 ؎ 0.6 mV), and there is a 10-fold increase in the percentage of channels that burst (fail to inactivate) during prolonged depolarization (0.025% S1885stop (n ‫؍‬ 7) versus 0.0028% wild type (n ‫؍‬ 9), p < 0.005). These results suggest that the charged structured region of the SCN5A C terminus plays a major role in channel inactivation, stabilizing the inactivated state.
Voltage-gated Na ϩ channels are integral membrane proteins (1,2) that not only underlie excitation in excitable cells but determine the vulnerability of the heart to dysfunctional rhythm by controlling the number of channels available to conduct inward Na ϩ movement (3). The Na ϩ channel ␣ subunit, which forms the ion-conducting pore and contains channel gating components, consists of four homologous domains (I-IV) (4). Each domain contains six ␣-helical transmembrane repeats (S1-S6), for which mutagenesis studies have revealed key functional roles (5). Na ϩ channel inactivation is due to rapid block of the inner mouth of the channel pore by the cytoplasmic linker between domains III and IV that occurs within milliseconds of membrane depolarization (6 -9). NMR analysis of this inactivation linker (gate) in solution has revealed a rigid helical structure that is positioned such that it can block the pore, providing a structural explanation of the functional studies (10). Inherited mutations of the III/IV linker in the cardiac Na ϩ channel disrupt normal fast inactivation and cause cardiac rhythm disturbances in the Long QT syndrome (11)(12)(13).
Subsequent analysis of additional Na ϩ channel mutations linked both to Long QT syndrome and another inherited arrhythmia, the Brugada syndrome, has revealed a critical and unexpected role of the C-terminal tail of the channel in the control of inactivation (14 -19). Point mutations in the C terminus shift the voltage dependence of inactivation, change the kinetics of both the onset of and recovery from inactivation, alter drug-channel interactions, and reduce entry of channels into an absorbing inactivated state (14,17,20,21).
Here we report a structural analysis of the C-terminal tail of the cardiac Na ϩ channel that sheds new light on mechanisms that control inactivation gating of this channel. Homology modeling of the C terminus, assuming similarity to the N-terminal domain of calmodulin, predicted that the C terminus would adopt a predominantly ␣-helical structure, a prediction verified by CD of a purified C terminus fusion protein. Functional studies revealed that only the proximal region of the C terminus, which contains all of the helical structure, markedly modulates channel inactivation but not activation. The distal Cterminal tail, which is largely unstructured, does not affect channel gating but affects the density of functional Na ϩ channels in the surface membrane. Our results suggest interactions between the structured region of the C terminus and other components of the channel protein that act to stabilize the channel in a pore-blocked inactivated state during membrane depolarization.

EXPERIMENTAL PROCEDURES
Materials-Unless otherwise noted, chemicals and reagents were obtained from Fisher or Sigma.
Subcloning/Expression-Oligonucleotide primers used to make PCR products for the C terminus (of SCN5A) expression systems are as follows (restriction sites underlined): 5Ј-primer (sense), GCATACGTC-CATATGGAGAACTTCAGCGTGGC, sequence beginning with SCN5A Glu 1773 including an NdeI restriction site; 3Ј-primer (antisense), CGT-CAGCTCGAGTCACACGATGGACTCACG, sequence ending with SCN5A C-terminal Val 2016 ; and CGTCAGCTCGAGTCAAGAGCGTTG-CAG, ending with Ser 1920 , including an XhoI restriction site and upstream stop codons. The NdeI/XhoI nucleotide segments were subcloned into corresponding sites of pET-28a(ϩ) (Novagen, Madison, WI). These constructs synthesize the protein with 21 additional amino acids at the N termini, including a poly(His) tag and a thrombin cleavage site.
Purification-Bacterial pellets were resuspended and lysed in a denaturing buffer (8 M urea, 100 mM NaH 2 PO 4 , 10 mM Tris, pH 8.0), 4 ml of buffer/g of wet pellet, rocking at room temperature for 2 h. The solution was spun at 12,000 ϫ g for 25 min at 21°C. The supernatant was retrieved and added to a Ni 2ϩ -nitrilotriacetic acid resin slurry (Novagen), 1 ml of resin slurry/4 ml of lysate, and rocked at room temperature for 1 h. Resin was washed twice with 1 lysate volume with 8 M urea buffer (8 M urea, 100 mM NaH 2 PO 4 , 10 mM Tris, pH 6.3) and spun at 1800 ϫ g for 2 min at 21°C, and supernatant was discarded. The wash protocol was repeated with 1 lysate volume each of 4, 2, and 0 M urea buffers (urea, 100 mM NaH 2 PO 4 , 10 mM Tris, 0.1% laryl sulfate, 10% ethanol, pH 8.0). Beads were resuspended in 1 lysate volume of elution buffer (100 mM NaH 2 PO 4 , 10 mM Tris, 0.05% laryl sulfate, 5% ethanol, 250 mM imidazole, pH 8.0) and rocked for 1 h at 21°C. Beads were spun, and supernatant was recovered. Purified protein was dialyzed into a 0 mM imidazole buffer for biophysical analysis.
Circular Dichroism-Purified proteins were prepared as above. CD spectra were measured from 200 to 250 nm using a Jasco J-810 spectropolarimeter (Easton, MD). All measurements were carried out at 22°C using a 0.1-mm path length quartz cuvette. Data were acquired at a resolution of 0.5 nm, bandwidth 1 nm. Eight scans were averaged at 100 nm/min, and buffer spectra were subtracted from the raw data.
Computational Analysis-A sequence homology search was done with the PSI-BLAST alignment program (22). The C terminus sequence was first used as a query to build the position-specific score matrix on the NCBI nr sequence data base with four iterations. The low complex-ity regions of the nr sequences have been masked using the SEG program with default parameters (23). The checkpoint file was then used in a following one-iteration run of the PSI-BLAST on the Protein Data Bank. Two families of proteins in the Protein Data Bank were identified as possible templates for the modeling of the three-dimensional structure of the first half of the C terminus peptide (from residue 1785 to 1885): the N-terminal domains of calmodulin and troponin C. These two structural domains are sequence-related and have essentially identical folding topology. The N-terminal domain of calmodulin (Protein Data Bank code 1vrk, chain A, residues 4 -95 (24)) is related to the C terminus peptide with a p value of 10 Ϫ8 , and the N-terminal domain of the troponin C (Protein Data Bank code 1a2x, chain A, residues 11-105 (25)) is related to the C-terminal peptide with a p value of 10 Ϫ6 . Previous work has demonstrated that protein pairs with p values less that 10 Ϫ6 are highly likely to adopt the same fold (26). We thus have compared the query sequence with more than 10,000 protein structures before selecting the appropriate template for modeling (27,28). Hence, we use the N-terminal domain of the calmodulin 1vrk (chain A residues 4 -95) as the template to build a model structure for the first half of the C terminus peptide (residues 1785-1885). Although the template structure has two EF hands for calcium binding, many of the key acidic residues that are used for specific calcium binding in calmodulin are not conserved in the C terminus peptide, suggesting that the C terminus peptide does not bind to calcium as in calmodulin. The structure of the second half of the C terminus peptide is not known. Nevertheless, the secondary structure predictions have suggested that the structure of the second half of the C terminus peptide adopts a long helical structure plus a long tail of nonstructural region.
Electrophysiology-Methods for mutagenesis of the human heart Na ϩ channel ␣ subunit, transient transfection of HEK 293, and patch clamp analysis of expressed channels were as described in previous publications (29). Single channel currents were recorded using the cell-attached configuration of the patch clamp with the following bath FIG. 2. Human sodium channel homology. Sequence alignment of the SCN5AC terminus and the C terminus of four other human voltage-gated sodium channels: SCN2A, the brain isoform; SCN8A, also from the central nervous system; SCN9A, the peripheral nervous system; and SCN4A, found in skeletal muscle. Colored letters distinguish homology: identical (red), conserved (blue), similar (purple), weakly similar (green), and nonhomologous (black). Alignment gaps are denoted by a dash. The six computed C terminus helices are shown by shaded bars below the sequences. Conserved domains and motifs are surrounded by boxes and labeled accordingly.
FIG. 1. Predicted structure of the C terminus of SCN5A. A, query-template alignment and the secondary structure predictions for the C terminus SCN5A. The first row is the sequence of the template, the N-terminal domain of calmodulin, Protein Data Bank code 1vrk:A; the second row shows the secondary structure assignment of the template structure using the DSSP program (46) (A, a, B, and b represent exposed ␣, buried ␣, exposed ␤, and buried ␤ residues, respectively); the third row is the sequence of the C terminus SCN5A; the fourth and fifth rows are the secondary structure predictions for the C terminus sequence (H, E, and C refer to ␣-helix, ␤-sheet, and random coil, respectively). B, predicted structure of the first-half of the C terminus SCN5A (residues 1785-1885). The backbone structure is shown with Asp residues in red and Glu residues in magenta. The sequence regions assigned to H1-H6 are shown in Fig. 2. C, schematic of the ␣ subunit of the channel, based on hydropathy plots. The protein consists of four domains each comprised of six transmembrane helices and a pore-forming region. The fourth helix (red) in each domain contains a series of positively charged residues and is the putative voltage sensor required for activation. The putative inactivation gate consisting of hydrophobic residues (IFM) is indicated by the circled h (yellow). The C-terminal region is detailed to give relative size and sequence position of the six predicted ␣-helices, H1-H6. solution: 140 mM KCl, 5 mM HEPES, 1 mM MgCl 2 (pH 7.4). Electrode resistance was typically 5-7 megaohms when filled with internal solution (110 mM NaCl, 10 mM HEPES, pH 7.4). After establishing the cell-attached configuration (seal resistance Ͼ10 gigaohms), the membrane was held at a holding potential of Ϫ120 mV. Test pulses (Ϫ30 mV, 100 ms) were applied every 0.5 s. The probability of bursting (Pb) is estimated by the equation, Pb ϭ (1 Ϫ (nb/t) 1/n ) ϫ 100, where nb, t, and n represent the number of nonbursting sweeps and total sweeps and channels, respectively. Sweeps with long lasting and repetitive opening activity are defined as bursting sweeps.

RESULTS
Theoretical Structure-We used the sequence of the C terminus of SCN5A for homology modeling. Since this section of the sodium channel is naturally present intracellularly (5), we assumed that it would fold similarly to proteins that are entirely cytoplasmic. Fig. 1A shows the PSI-BLAST alignment of the query sequence to the template (see "Experimental Procedures"). Secondary structure predictions of the C terminus peptide in Fig. 1A were obtained from the servers of PSI-PRED (30) and PHD (31). The predicted structure of the first half of the C terminus domain consists of six helices (H1-H6; Fig. 2). The three-dimensional model of the first half of the C terminus peptide is shown in Fig. 1B. Only the first five helices' three-dimensional locations can be predicted. Although the sixth, long helix is present, its relative location cannot be calculated with accuracy.
The tertiary structure of this region of the channel (Fig. 1B) has been modeled with PrISM (32). As expected for a cytoplasmic protein, hydrophilic residues are generally facing the outside of the structure, while hydrophobic amino acids are protected in the globular core. One particularly interesting property of this globular structure is that it has a density of acidic residues on one end (H1 and H2) and basic residues, primarily in helix H6, that are likely to be distributed on the opposite side of the structure to form a bipolar arrangement of the real charges in the structure. However, since the predicted structure does not reveal the exact positions of the basic residues in the structure, the hypothesis of the charge distribution remains to be tested. Sequence comparison between SCN5A and other human sodium channel isoforms (Fig. 2) reveals that those regions responsible for this suggested polarity are highly conserved, particularly the acidic rich domain, containing H1 and H2. On the other hand, the sequences of the distal half of the C-terminal domain after H6 are diverse with less conserved sequence features and extensive insertions and deletions. This is in agreement with the expectation that the distal half of the protein is not structured. If the functional properties of the C terminus are structurally dependent, then the distal half of the tail is not likely to be involved in channel gating (see below). Fig. 1C illustrates, in schematic format, the relationship be- FIG. 3. Purification of SCN5AC terminus. Two C terminus peptides were expressed in BL21(DE3) cells, with N-terminal His tags, and purified over a nickel affinity column. Coomassie stain of the purified proteins is seen in A. Lane 1, the full-length construct; lane 2, the proximal half of the C terminus. Protein identity was confirmed by Western blot using two separate antibodies: one against the Histag and the other against the C-terminal 17 amino acids of SCN5A (schematic). His-tagged proteins are shown in B, and those containing the C terminus of SCN5A are shown in C. The shorter construct is not recognized by the sodium channel antibody, because it does not contain the residues against which the antibody was raised.
FIG. 4. CD spectra of C terminus regions of SCN5A. Comparison of the C terminus constructs' CD spectra. Each spectrum is represented as the mean residue molar ellipticity. The spectrum of the predicted structural region (residues 1774 -1885, closed squares) shows more structural components, particularly ␣-helical content, relative to the full C terminus (residues 1774 -2016, open squares), as seen by the increase in the ellipticity at 222 nm. The spectra were obtained by the subtraction of the buffer spectra from that of the sample. Protein Purification-Two constructs of the C terminus of SCN5A were made and subcloned into an expression vector containing a T7 viral promoter. One construct contained the entire length of the C terminus, residues 1774 -2016, and the other included simply the proximal half, the region predicted to be enriched in structure, residues 1774 -1920. Expressed in bacterial cells containing a viral polymerase transposon, the T7 promoter is constitutively transcribed when induced by isopropyl-␤-D-thiogalactoside. We included an N-terminal His tag to be used for affinity purification over a nickel column. The constructs were difficult to purify under native conditions, due to insolubility of the protein as well as its sensitivity to proteolytic degradation, particularly in the case of the full-length construct. Consequently, bacterial pellets were lysed under denaturing conditions and progressively washed into more native solutions while bound to the nickel resin. This gave high yields of purified protein in solution. Protein identity was con-firmed by Western analysis with both antibodies raised against the poly(His) tag as well as the C-terminal 17 amino acids of SCN5A (Fig. 3). The purification of recombinant proteins from insoluble pellets under denaturing conditions has been used previously in studies in which structures have been solved through refolding, including the glutamate receptor (33), a membrane protein. However, in the case of the full-length construct for the SCN5A C terminus, it was also possible to purify a very small amount of protein from the soluble fraction. CD spectra of these proteins were obtained and found to be consistent with those of the refolded proteins (data not shown). Therefore, proteins purified using the refolding method were used in the following experiments.
Circular Dichroism-Purified proteins were dialyzed to buffer without imidazole, since the compound is known to rotate light in the ultraviolet range. Each protein spectrum is the product of the average of eight scans minus the spectrum of the buffer solution. The CD spectra for the two constructs are shown in Fig. 4. The full-length C terminus (Fig. 4, open   FIG. 5. Influence of C-terminal truncations on inactivation gating. A, families of whole cell currents in response to a series of 25-ms voltage pulses (Ϫ60, Ϫ25, Ϫ10, 0, and ϩ20 mV) along with schematic representation of ␣ subunit constructs. Truncation of H6 along with the distal C terminus (S18885stop) results in sustained channel activity (arrow) at the end of deoplarizing pulses. Holding potential was Ϫ100 mV. Shown are averaged and normalized currents (WT, n ϭ 7; L1921stop, n ϭ 7; S1885stop, n ϭ 9). B, peak was measured at each voltage, normalized to maximal current over all voltages, averaged, and plotted Ϯ S.E. versus test voltage (⅜, WT, n ϭ 7; OE, S1885stop, n ϭ 7; f, L1921stop, n ϭ 9). C, influence of C-terminal truncation on the voltage dependence of steady-state inactivation and activation. Inactivation was measured as the influence of conditioning pulses (500 ms; Ϫ120 to Ϫ20 mV) on peak current measured at Ϫ10 mV (test pulse). Test pulse current was normalized to the current measured after the Ϫ120 mV conditioning pulse and plotted versus the conditioning voltage for each construct (E, WT; ‚, S1885stop; Ⅺ, L1921stop). The voltage dependence of activation was measured by normalizing peak test pulse (25 ms; Ϫ80 to ϩ50 mV) current amplitude to driving force defined as the difference between estimated reversal potential in B (ϩ42 mV) and test pulse voltage. (q, WT; OE, S1885 stop ; ;, L1921 stop ). To ensure voltage control, the external solution contained a reduced concentration of Na ϩ (30 mM) (see "Experimental Procedures"). Experimental data were fitted with Boltzmann relationships to obtain the following parameters. For inactivation, V1 ⁄2 is Ϫ61.9 Ϯ 0.2 mV (WT, n ϭ 8), Ϫ73.2 Ϯ 0.6 mV (S1885stop, p Ͻ 0.001, n ϭ 12), and Ϫ63.3 Ϯ 0.5 mV (L1921stop, not significantly different from WT, n ϭ 6). The slope factor k is 6 Ϯ 0.14 (WT), 5.8 Ϯ 0.1 (S1885stop, not significantly different from WT), and 5.5 Ϯ 0.13 (L1921 stop, not significantly different from WT). For activation, V1 ⁄2 is Ϫ22.8 Ϯ 0.5 mV (WT, n ϭ 8), Ϫ21.9 Ϯ 0.9 (S1885stop, not significantly different from WT, n ϭ 9), and Ϫ21.5 Ϯ 0.8 (L1921stop, not significantly different from WT, n ϭ 7). The slope factor k is 7.2 Ϯ 0.3 (WT), 7.9 Ϯ 0.3 (S1885stop, not significantly different from WT), and 8.1 Ϯ 0.5 (L1921 stop, not significantly different from WT). squares) exhibits some definitive structure; however, it is not clearly dominated by any of the known secondary structures. As predicted by the computational model, the CD of the proximal half of the C terminus (Fig. 4, closed squares) is more clearly ordered and is very similar to spectra of all-helical proteins (known to have two distinct negative peaks at 208 and 222 nm (34)). In order to estimate the relative secondary structure content, we employed two computer algorithms, CON-TINLL and CDSSTR (35,36). The two programs showed the full-length construct to be composed of an even mixture of ␣-helix and nonstructure, while the shorter construct is almost two-thirds ␣-helical content (Table I). These data provide biophysical evidence that supports the predicted structure of the C terminus; the proximal half is predominantly helical, and the distal half is largely unstructured.
Functional Roles of C Terminus Structures: Electrophysiological Experiments-To further study the predictions of the model, we investigated the functional consequences of cleaving 1) a predicted nonstructured segment (construct L1921stop) and 2) a predicted helical portion of the C terminus along with the above unstructured region (S1885stop) using electrophysiological properties as an assay. The L1921stop mutation leaves the entire predicted structural region intact, while removing the nonstructured distal portion of the tail, and corresponds to the shorter construct studied above. Biophysical properties of channels encoded by this construct are similar to those encoded by full-length (WT) 1 ␣ subunits (Fig. 5), not differing significantly from wild type in any observed property of the channel, with the exception of reduced peak current (Ϫ180.07 Ϯ 19.7 pA/pF (n ϭ 17) versus Ϫ403.18 Ϯ 43.0 pA/pF (n ϭ 21), respectively, p Ͻ 0.05). This rules out the role of the distal portion of the C terminus in channel gating. The L1885stop mutation, which cleaves positively charged residues between amino acids 1885 and 1921 and the predicted sixth helix, reduces peak currents (Ϫ266.17 Ϯ 17.3 pA/pF (n ϭ 37) versus Ϫ403.18 Ϯ 43.0 pA/pF (n ϭ 21) for WT; p Ͻ 0.05) but, in addition, changes channel gating. The S1885stop mutation shifts steady state inactivation (Ϫ73.2 Ϯ 0.6 mV (n ϭ 12) for mutant versus Ϫ61.9 Ϯ 0.2 mV (n ϭ 8) for WT, p Ͻ 0.001) and slows the recovery from inactivation (time to half-recovery 3.01 Ϯ 0.06 ms for WT (n ϭ 8); 5.92 Ϯ 0.06 ms for S1885stop (n ϭ 5), p Ͻ 0.05; not shown). Perhaps most striking, however, is the pronounced effect of this mutation on Na ϩ channel currents measured during prolonged depolarization (Fig. 5, arrow, S1885stop traces). As can be seen more clearly in the inset of Fig. 6A and in the summary graph of Fig. 6C, this truncation induces a significant increase in sustained channel activity compared with both WT and L1921stop channels (1.78 Ϯ 0.2 pF/pA (n ϭ 24) for mutant versus 0.24 Ϯ 0.05 pF/pA (n ϭ 14) for WT, p Ͻ 0.001; 0.27 Ϯ 0.07 pA/pF (n ϭ 5), L1921stop), suggesting in- 1 The abbreviations used are: WT, wild type; pA/pF, picoamps/ picofarads. creased bursting activity in these channels. Single channel experiments reveal that this is the case (Fig. 7). Shown in Fig.  7 are consecutive current traces showing activity of a small number (10) of channels in cell-attached membrane patches. Recordings from WT channels reveal few reopenings during prolonged depolarization, consistent with channels entering an absorbing inactivated state. Infrequently, however, reopenings are observed for WT channels, reflecting a low probability of entrance into a bursting mode of channel activity, even with the full-length C-terminal tail. However, bursting is observed much more frequently in the case of recordings from S1885stop channels (Fig. 7), causing a significant increase in the probability that channels burst when compared in a large number of experiments (Fig. 7). This type of channel activity, shown to be arrhythmogenic in inherited cardiac disorders (11), indicates that a significant fraction of channels no longer enter an absorbing inactivated (nonconducting) state during prolonged depolarization and underlies sustained whole cell channel activity illustrated in Figs. 5 and 6. While only a fraction of the peak channel activity, this small fraction of noninactivating channels can conduct currents that are sufficient to prolong cellular action potentials and trigger fatal cardiac arrhythmias (37). DISCUSSION We have combined theoretical modeling, CD measurements, and functional expression studies to probe the structure and functional role of the C terminus of the human voltage-gated Na ϩ channel ␣ subunit in the control of channel gating. CD measurements provide strong support for the computational predictions of the C terminus structure, and the functional studies provide new insights into mechanisms by which these structures may contribute to the control of channel gating.
The agreement between theoretical prediction and experimental measurement of the C terminus structure is strong. The CD data show that the first half of the protein is well structured and mostly helical, and the distal half of the protein is largely unstructured, supporting the predictions of the homology modeling. Because the distal half does not affect either activation or inactivation gating, our results suggest that without coordinated structure, this region of the C terminus does not interact with other channel components responsible for gating. On the other hand, the computational work suggests that the proximal half of the protein adopts a fold similar to calmodulin with the charged residues distributed in a bipolar pattern. Thus, the computational work predicts that the proximal helices may play important roles in controlling channel gating.
The predicted structure reveals several observations regarding the charge distribution within the C terminus. First, similarly charged residues in general occupy positions on a common side of helices. This is particularly true in H1 and H2 as FIG. 7. Truncation of C terminus helix 6 (H6) promotes bursting of Na ؉ channels. Shown are cell-attached patch recordings of WT and S1885stop mutant channels in response to prolonged (150-ms) pulses to Ϫ10 mV. Typical consecutive (five) sweeps are shown in each panel. Each patch contained 10 channels. The dashed line indicates base-line current level with no Na ϩ channels open. The bar graph plots bursting frequency (Pb) (see "Experimental Procedures") for both WT and S1885stop channels. Data for bursting frequency were obtained from nine cells for WT patches and seven cells for S1885stop patches. **, p Ͻ 0.005. well as H6. Second, although residues of both positive and negative charge exist throughout the C terminus, helices H1 and H2 (residues 1780 -1820) have a high concentration (14 of 41) of negatively charged amino acids, and H6 has a high concentration (7 of 27) of positively charged residues. Together, and depending on the precise localization of H6, this may provide for the formation of a bipolar globular structure, which may show functional significance.
Functional studies provide support for this speculation and suggest that C-terminal structures play an important and specific role in controlling channel inactivation. In our experiments, truncation (S1885stop mutation) of the model-predicted sixth helix, which contains positive charges in the end of the proximal half of the C terminus, caused significant changes in inactivation but not activation gating. Most notable was a significant increase in the fraction of channels that failed to enter an absorbing inactivated state during maintained depolarization. This fraction of channels is evident as the mutationinduced increase in maintained current in Fig. 6 and the increase in bursts of single channel activity seen in Fig. 7. These changes in inactivation occur despite the fact that the voltage dependence of activation of the channel is not affected. In contrast, however, the voltage dependence of inactivation is shifted toward more negative potentials by the S1885stop mutation. This means that less energy is expended (more modest depolarization) in reducing the availability (increasing the fraction of closed state inactivated channels). Additionally, this truncation slows the recovery of channel from the inactivated state by hyperpolarization. These latter effects of the truncation of H6 are consistent with stabilization of the closed inactivated state. Thus, it appears that these experiments reveal a complex role of the C terminus helices in control of inactivation possibly by independently influencing protein conformations that underlie transitions of modal activity (bursting) of the channel and rapid gating (inactivation).
Additional evidence for the importance of the proximal region of the C terminus in the control of channel inactivation comes from the analysis of inherited mutations of the Na ϩ channel ␣ subunit that are linked to the cardiac rhythm disturbances, Long QT syndrome and Brugada syndrome. Within the last few years, several mutations linked to these diseases have been reported in this region of the Na ϩ channel ␣ subunit, which has been referred to as an "acid-rich" domain (18). These include mutations E1784K (18), D1795insD (16,17), Y1795H, Y1795C (38), and D1790G (14,15,20). All of these inherited mutations are in the acid-rich first five helices of the proximal region of the C terminus; all cause changes in channel inactivation with no reported effects on channel activation; and all disrupt normal cardiac rhythm and place mutation carriers at risk for sudden cardiac death.
Fast inactivation of Na ϩ channels is coupled to activation of the channels (39). Transitions into the inactivated state can occur following activation or opening of the channel that occurs when charged S4 segments of the four domains of the ␣ subunit respond to membrane depolarization (40,41). Inactivation of open channels develops rapidly as the intracellular peptide that links domains III and IV occludes the channel by folding and binding to a receptor in the intracellular mouth of the pore, preventing ion flow (6,(42)(43)(44) similar to "ball and chain" inactivation of K ϩ channel inactivation (45). Three-dimensional structural analysis of the central portion of the inactivation gate as well as mutagenesis studies support the view that the inactivation gate forms a hydrophobic interaction with its receptor to occlude the pore (10, 42).
Because we find that the proximal structured and charged region of the C terminus specifically modulates channel inactivation, our findings suggest that electrostatic interactions, provided by the specific distribution of the charged residues on a well defined structure, may contribute an energetic source in modulating the inactivation gating mechanism, in addition to the ball-and-chain type mechanism, where specific short range hydrophobic contacts are the major driving force in the inactivation mechanism (42,43). The significance of the findings is that mechanisms with completely different energetic origin could exist in parallel in controlling the gating of the channels. Elucidating the details of the mechanisms requires further experimental and computational work.