Solution Structure of the NaV1.2 C-terminal EF-hand Domain*

Voltage-gated sodium channels initiate the rapid upstroke of action potentials in many excitable tissues. Mutations within intracellular C-terminal sequences of specific channels underlie a diverse set of channelopathies, including cardiac arrhythmias and epilepsy syndromes. The three-dimensional structure of the C-terminal residues 1777-1882 of the human NaV1.2 voltage-gated sodium channel has been determined in solution by NMR spectroscopy at pH 7.4 and 290.5 K. The ordered structure extends from residues Leu-1790 to Glu-1868 and is composed of four α-helices separated by two short anti-parallel β-strands; a less well defined helical region extends from residue Ser-1869 to Arg-1882, and a disordered N-terminal region encompasses residues 1777-1789. Although the structure has the overall architecture of a paired EF-hand domain, the NaV1.2 C-terminal domain does not bind Ca2+ through the canonical EF-hand loops, as evidenced by monitoring 1H,15N chemical shifts during aCa2+ titration. Backbone chemical shift resonance assignments and Ca2+ titration also were performed for the NaV1.5 (1773-1878) isoform, demonstrating similar secondary structure architecture and the absence of Ca2+ binding by the EF-hand loops. Clinically significant mutations identified in the C-terminal region of NaV1 sodium channels cluster in the helix I-IV interface and the helix II-III interhelical segment or in helices III and IV of the NaV1.2 (1777-1882) structure.

Voltage-gated sodium channels (VGSCs) 5 are molecular assemblies that span the plasma membrane of excitable cells and conduct sodium current selectively in response to depolarizing stimuli. Mutations in VGSCs underlie a variety of diseases, including the cardiac arrhythmogenic Long-QT3 and Brugada syndromes (1,2) and neurological syndromes, such as epilepsy (3,4).
Known components of VGSCs include a pore-forming ␣ subunit, auxiliary ␤ subunits, and associated modulating proteins, such as calmodulin (5,6). The ␣ subunit is composed of four homologous six-transmembrane helical domains connected by inter-domain linkers and N-terminal and C-terminal cytoplasmic regions. Specific ␣ subunit isoforms are expressed differentially in skeletal muscle (Na V 1.4), cardiac muscle (Na V 1.5) and the nervous system (Na V 1.1, Na V 1.2, Na V 1.3, splice variants of Na V 1.5, and Na V 1.6-Na V 1.9) and control the rapid upstroke of action potentials (7). VGSC activity is characterized by two open states and several inactivated states (8). Kinetics of channel inactivation occur on timescales ranging from milliseconds to seconds and determine multiple aspects of action potentials (9,10). The molecular mechanisms of VGSC inactivation are complex and involve the ␣ subunit, the ␤ subunits, and calmodulin (11)(12)(13). Specific contributions to ␣ subunit inactivation have been localized to interhelical intra-domain regions (14 -16), the linker region between domains III-IV, which forms the pore occluding inactivation gate (17,18), and the C-terminal cytoplasmic domain (CTD) (19 -21).
Specific disease-causing mutations within the CTD affect channel function by altering kinetics of channel inactivation (22). The CTD is predicted by sequence analysis (23,24) and homology modeling (25)(26)(27) to contain a paired EF-hand domain and was observed to contain a distal calmodulin binding IQ motif (4, 12, 28 -31). Structural modeling also predicts that specific interactions between helix I and helix IV control channel inactivation (27,32). A recent model, based on NMR chemical shift perturbations, fluorescence spectroscopy, and electrophysiology, suggests that inactivation is regulated by Ca 2ϩ binding to the proximal EF-hand, which is strongly influenced in turn by interactions with the distal IQ motif and cal-modulin (33). Nevertheless, whether Ca 2ϩ binds specifically to the putative CTD EF-hand and any resultant contribution to channel regulation is controversial (12,26,31,34).

EXPERIMENTAL PROCEDURES
Constructs of the Na v 1.2 CTD were designed by limited proteolysis and H/D exchange experiments. Briefly, the CTD of Na v 1.2, residues 1777-1937 with the amino acid substitutions I1877A/Q1878A and an N-terminal His 6 tag MGSSHHHH-HHSSGLVPRGSHMAS (31), was subjected to proteolytic digestion with proteinase K at 4°C for 15-60 min using a protein:protease ratio of 50:1-100:1. The termini of the protected proteolytic fragments were mapped by matrix-assisted laser desorption ionization time-of-flight time-of-flight mass spectrometry and N-terminal sequencing. H/D exchange experiments were performed by ExSAR (Monmouth Junction, NJ) and showed protection for proteolytic fragments extending from residues 1789 to 1879. The construct encompassing residues 1777-1882 of the Na v 1.2 CTD defined by the above experiments, including the N-terminal His tag, was used for structure determination by solution NMR spectroscopy.
An initial structure of Na V 1.2 CTD was calculated from dihedral angle and NOE distance restraints, with several iterations to resolve ambiguity using ARIA 2.2 (45) and CNS 1.2 (46). The structure was refined with XPLOR-NIH 2.18 using dihedral angle, NOE distance, and residual dipolar coupling constants restraints (47,48). Dihedral angle restraints were derived from chemical shifts using TALOS (49). Distance restraints were obtained from NOE intensities corrected for multiplicity of the 1 H spins. NOE connectivities were categorized into three classes (50). Class I contains all intra-residue H N -H ␣ and all intra-residue, sequential, and medium range H ␤ -H X NOEs, where X is not a methyl proton. Class III contains all NOEs involving a methyl group, and class II includes all other NOEs. A calibration factor, k I , was obtained by equating the average class I intensity to 3.4 Å. The class II calibration factor k II ϭ k I /2.4 2 . The class III calibration factor k III ϭ k II /2. Class I was averaged with a 1 ⁄ 6 order exponent, whereas classes II and III were averaged using a 1 ⁄ 4 exponent (50,51). A standard 10% error term was applied to the upper bound of each restraint. All distances were constrained to the range (1.8, 5.5 Å). Pseudo atom corrections were applied to upper distance restraints for geometric considerations (52).
The 1 H, 15 N, 15 N-13 CЈ, and 1 H ␣ -13 C ␣ residual dipolar coupling constants were included in the structure calculations. The residual dipolar coupling magnitude and rhombicity were set to Ϫ12.5 Hz and 0.55, respectively, during the initial minimization and were refined in the final all-atom minimization. The final average residual dipolar coupling magnitude and rhombicity are Ϫ12.8 Ϯ 0.23 Hz and Rh ϭ 0.56 Ϯ 0.01, respectively, for the 200 conformers.
Structural quality statistics refer to residues Leu-1790 -Glu-1868 of the 15 lowest-energy structures of 200 total structures calculated. NOE completeness was determined with aqua3.2 (53). The Pearson correlation coefficient (R) and the quality factor (Q) were computed with PALES (54) from 64 CЈ-C ␣ dipolar couplings that were not included in the structure calculation. MolProbity scores were calculated for the lowest energy structure (55). Average root mean square deviation values were calculated to the average coordinates with VMD (56). Interhelical distances and angles (rounded to the nearest degree) were computed using interhlx. 6 Structural alignments were performed with CE (57), and structure figures were prepared with VMD (56) and MOLMOL (58).

RESULTS
The isolated Na V 1.2 CTD (1777-1882) and Na V 1.5 CTD (1773-1878) constructs each contain the region just after their respective predicted IVS6 transmembrane helix and extend to a region highly conserved among all VGSCs just before the IQ motif. Assignments of 1 H, 15 N resonances for the Na V 1.2 CTD and the Na V 1.5 CTD are, respectively, 99 and 97% complete. Notably, Asn-1835 could not be assigned in the 1 H, 15 N HSQC of Na V 1.2. The resonances for Asn-1831 (the homologue of Asn-1835) and Gln-1832 were not assigned, and the resonance for Ile-1833 appears broadened in 1 H, 15 N HSQC of Na V 1.5. Moreover, homologous resonances Leu-1855 in Na V 1.2 and Met-1851 in Na V 1.5 have liminal intensities in 1 H, 15 N HSQC spectra. These observations suggest conserved dynamics between isoforms. For the Nav1.2 CTD (1777-1882), 13 C ␣ and 13 C ␤ assignments are 100% complete, 13 CЈ assignments are 97.1% complete, 1 H aromatic assignments are 89.1% complete, and non-aromatic 1 H assignments are 97.7% complete. The Na V 1.2 CTD construct contains six proline residues, of which Pro-1789, Pro-1807, Pro-1827, and Pro-1845 are in a trans conformation, whereas Pro-1828 and Pro-1834 are in a cis conformation. The cis conformation is evidenced by stronger X-Pro H ␣ -H ␣ than X-Pro H ␣ -H ␦ NOE contacts and differences of C ␤ -C ␥ chemical shifts of 9.4 and 8.5 ppm, respectively (59, 60). Medium range 1 H-1 H NOEs, steady-state { 1 H}-15 N NOE, and 13 C ␣ secondary chemical shifts for Na V 1.2 indicate that the CTD forms a well folded domain between residues Leu-1790 and Glu-1868, with a less well ordered region between residues Ser-1869 and Arg-1882 and a disordered N-terminal region between residues Gly-1777 and Pro-1789 ( Fig. 1 and supplemental Fig. S1). Secondary chemical shifts indicate that the Na V 1.5 CTD has a similar secondary structural architecture as Na V 1.2 CTD (Fig. 1C).
The structure of Na V 1.2 CTD is presented in Fig. 2 and supplemental Fig. S1 with statistical details of the calculation presented in Table 1. The structure contains four ␣-helices and two short anti-parallel ␤-strands, consistent with homology models based on structures of paired EF-hand domains (25,26). Comparison of interhelical angles of helices I and II of Na V 1.2 CTD and the N-terminal lobe of the prototypical EF-hand protein calmodulin suggests that the isolated Na V 1.2 CTD most closely resembles the canonical apoEF-hand conformation ( Fig.  3D and Table 2). The hydrophobic interface between helices I and IV predicted through mutational analysis (27) is observed with direct NOE contacts between residues Phe-1795, Phe-1798, and Tyr-1799 in helix I and Leu-1855, Ile-1857, and Leu-1858 in helix IV.
Helices I and IV contribute to the hydrophobic core of the protein, with a majority of aromatic side chains contributed from helix I. The segments between Gln-1811-Glu-1814 and Arg-1851-His-1853 participate in an anti-parallel ␤-sheet. An additional anti-parallel ␤-sheet contribution from residues Met-1846 -Val-1847 is not present in all conformers of the structural ensemble. The helix II-III interhelical segment, delimited by two cis proline residues, Pro-1828 and Pro-1834, is well ordered in the structural ensemble. The conformation of residues Asp-1826 -Leu-1829 is consistent with a type VI tight-turn (61), also called a ␤␣ R turn (62). The unique di-proline-leucine motif, Pro-1827-Leu-1829 extends the helix II-III interhelical segment by forming a small handle at the base of helix II (Figs. 3, D and E). The absence of long-range NOE contacts for residues Ser-1848 and Gly-1849 is represented by disorder of this region in the ensemble.
The segment from residues Ser-1869 to Arg-1882 is predicted to have residual helical content based on secondary 13 C chemical shifts and characteristic d ␣N (i, iϩ3) and d ␣ ␤(i, iϩ3) NOEs (Fig. 1). A short helix V is observed in the final ensemble extending from Gly-1870 to Arg-1876 with a backbone root mean square deviation of 0.59 Å when superposed on itself (supplemental Fig. S1). However, the reduced magnitudes of the secondary 13 C chemical shifts and the { 1 H}-15 N NOEs for helix V compared with helices I to IV, suggest that the helical conformation is not fully populated in solution. Furthermore, helix V does not exhibit residual dipolar couplings or longrange NOE contacts and, hence, is not well defined relative to the core EF-hand domain structure (supplemental Fig. S1). Additional interactions present in longer constructs of the CTD or in complexes with other components of the VGSC may stabilize helix V.
Binding of Ca 2ϩ by the Na V 1.2 (1777-1882) and Na V 1.5 (1773-1878) CTDs was assessed by monitoring 1 H, 15 N chemical shifts as a function of Ca 2ϩ concentration (0 -4.5 mM). Chemical shift perturbations exhibit titration behavior suggesting that the interaction occurs on a fast-exchange timescale with equilibrium constants of 1.65 Ϯ 0.03 mM for Na V 1.2 CTD and 3.28 Ϯ 0.13 mM for Na V 1.5 CTD (Fig. 3 and supplemental Fig. S2), consistent with a previous report for the Na V 1.5 CTD (33). However, resonance assignments were not obtained previously, and the structure of Na V 1.2 CTD now reveals that chemical shift perturbations Ͼ0.05 ppm are localized to residues in the N terminus of helix I, the linker between helices II and III, the C terminus of helix IV and the partially structured helix V. Thus, this weak Ca 2ϩ binding site is distal to the canonical EF-hand loop motifs. In contrast, the average chemical shift change between the end points of the titration is Ͻ0.01 ppm in the N-terminal EF-hand loop (residues 1806 -1817) and in the C-terminal EF-hand loop (residues 1842-1853) for the Na V 1.2 CTD. Respective values Ͻ0.02 ppm were obtained for corresponding residues 1802-1813 and 1832-1849 in the Na V 1.5 CTD. In comparison, the average chemical shift changes of the N-terminal EF-hand loop between apoCa 2ϩ and Ca 2ϩ -loaded calmodulin are 0.59 and 0.65 ppm in the N-terminal and C-terminal domains, respectively (63,64). In particular, canonical Ca 2ϩ binding by an EF-hand would require coordination of a Ca 2ϩ atom by the backbone carbonyl atoms of Phe-1812 in Na V 1.2 and Phe-1808 in Na V 1.5, leading to significant chemical shift changes for inter-residual and sequential amide resonances (65,66). In opposition, chemical shift changes less than 0.02 ppm were observed for backbone amide resonances for residues Phe-1812-Ile-1813 and Phe-1808 -Ile-1809 of Na V 1.2 and Na V 1.5, respectively (Fig. 3). A structurebased sequence alignment of calmodulin and Na V 1.2 and a comparison of Ca 2ϩ -induced chemical shift changes are shown in supplemental Fig. S3.

DISCUSSION
The solution structure determined by NMR spectroscopy for the Na V 1.2 CTD (1777-1882) exhibits a core-ordered domain from residues Leu-1790 to Glu-1868, with four ␣-helices and two short anti-parallel ␤-strands arranged in tandem helixsheet-helix motifs characteristic of paired EF-hand domains.

Structure of the Na V 1.2 C-terminal EF-hand
Structural alignment of the Na V 1.2 CTD and calmodulin reveals that the structure is more similar to apo-Ca 2ϩ calmodulin than to peptide target and/or Ca 2ϩ -loaded calmodulin. The Na V 1.5 CTD (1773-1878), which shares 83% identity with the Na V 1.2 CTD, adopts a similar secondary structure and, likely, tertiary structure.
Titrations monitored by NMR chemical shift perturbations demonstrate that the canonical EF-hand loops of the Na V 1.2 CTD (1777-1882) and Na V 1.5 CTD (1773-1878) do not bind Ca 2ϩ ; rather, Ca 2ϩ binds weakly at a site distal to the canonical loops near the N terminus of helix I, the linker between helices II and III, the C terminus of helix IV, and the partially structured helix V. The high resolution crystal structure of calmodulin identified an additional Ca 2ϩ binding site in the homologous region corresponding to the helix II-III linker, but the authors judged this site to be non-physiological (67).
A structure-based sequence alignment with calmodulin also suggests that the canonical EF-hand loops of Na V 1.2 CTD do not bind Ca 2ϩ (Table 3  Lys residue present in Na V 1.2. Mutation of the corresponding residue, Glu to Lys, in Drosophila melanogaster calmodulin abolishes Ca 2ϩ binding, although this mutation may mimic a Ca 2ϩ -bound state in the context of certain targets (69,70). Lys is found at position 12 in the non-canonical Ca 2ϩ binding loop of scallop myosin essential light chain; however, coordination of Ca 2ϩ is accomplished by an acidic residue at position Ϫ2, the backbone carbonyl group at position ϩ2, and a water molecule (71). In Na V 1.2 the residue at position ϩ2 is Pro, and the residues at positions Ϫ3 and Ϫ2 are Glu and Lys. The latter two residues have chemical shift changes less than 0.05 ppm after the addition of 4.5 mM Ca 2ϩ .
Higher affinity Ca 2ϩ binding has been reported for longer constructs of Na V 1.5 CTD, residues 1773-1920 and residues  Table 1; the structure of the construct from residues Gly-1777 to Arg-1882, including the N-and C-terminal regions, is shown in supplemental Fig. S1. Traces through the backbone heavy atoms of the 15 lowest energy conformers described in Table 1

Structure of the Na V 1.2 C-terminal EF-hand
1773-1925 that include the IQ motif, and binding is abolished by mutation of the IQ motif (33). However, the resonance assignments obtained for Na V 1.5 indicate that chemical shift perturbations for key EF-hand canonical loop residues Phe-1808 -Ile-1809 are not larger in these longer constructs (comparing the inset of Fig. 3B with supplemental Fig. 5D of Ref. 33), suggesting that higher affinity binding of Ca 2ϩ also does not involve the canonical EF-hand loops.
The solution structure of Na V 1.2 CTD can be used to predict the effect(s) of clinical mutations in VGSCs (Fig. 4) because of the high degree of homology between VGSC CTDs. Generally, clinically significant mutations that map in the CTD can be divided into two classes, with some overlap for several sites (supplemental Table SI). Mutations in Na v 1.5 associated with the Long QT variant 3 (LQT3) cardiac arrhythmia phenotype and a subset of mutations in Na v 1.1 associated with certain epilepsy syndromes lead to persistent current during maintained depolarization. A second set of mutations in Na v 1.1 associated with multiple epilepsy syndromes and mutations in Na v 1.5 associated with the Brugada syndrome cardiac arrhythmia led to decreased current, resulting from loss of function or enhanced inactivation kinetics.
Multiple mutations in Na V 1.1 and Na V 1.5 associated with an increased persistent current are observed at positions clustering in the corresponding helix I of the Na V 1.2 CTD. The F1808L  Structure of the Na V 1.2 C-terminal EF-hand MARCH 6, 2009 • VOLUME 284 • NUMBER 10 mutation associated with intractable childhood epilepsy with generalized tonic clonic seizures in Na V 1.1 may destabilize the protein core because the aromatic ring of Phe-1798 in Na V 1.2 contacts residues in helix IV and the helix II-III interhelical segment (4,72). The insertion of an Asp residue at position 1795, Y1795insD, leads to both LQT3 and Brugada syndrome phenotypes in Na V 1.5 and potentially disrupts helix I by shifting the register of helical interactions (73). Substitution at position Tyr-1795 in Na V 1.5 differentially leads to decreased inactivation for Y1795C in LQT3 or enhanced inactivation kinetics for Y1795H in Brugada syndrome, whereas both substitutions lead to sustained current during maintained depolarization and negative shift of voltage dependence of inactivation (27,74). The Y1795C mutation has been suggested to form an intra-molecular disulfide bond with Cys-1850 in Na V 1.5 (32). The average C ␤ -C ␤ distance of the corresponding residues in the Na V 1.2 CTD structural ensemble is 9.6 Ϯ 0.4 Å. The C ␤ -C ␤ distance in cysteine disulfide bonds ranges from 3.4 to 4 Å (75); thus, the proposed disulfide bond may be intermolecular or require structural rearrangement on the order of several angstroms between helix I and IV (Fig. 4) if it is formed. Furthermore, although Tyr-1795 in Na V 1.5 was predicted to contribute to the hydrophobic interface between helices I and IV (27), the corresponding residue Tyr-1799 in

Structure of the Na V 1.2 C-terminal EF-hand
Na V 1.2 is found in a position closer to the surface; the total side-chain exposed surface area is 103 Ϯ 10 Å 2 for the conformers in Table 1. Hence, mutations at position Tyr-1799 may also affect interactions with other components of the intact channel.
On the other hand, the conserved Trp-1802, corresponding to Trp-1798 in Nav1.5, is not completely accessible as observed previously (27); the total side-chain exposed surface area is 9 Ϯ 5 Å 2 for the conformers in Table 1.
The L1825P mutation associated with LQT3 and the R1826H mutation associated with sudden infant death syndrome in Na V 1.5 occurs in the helix II-III interhelical segment (76,77). The L1825P mutation results in significant persistent current and slows kinetics of inactivation. Interestingly, the L1825P mutation in Na V 1.5 introduces a di-proline motif, as is observed in wild type Na V 1.1, Na V 1.2, Na V 1.3, and Na V 1.7, but shifted by one residue. The residue corresponding to Arg-1826 in Na V 1.2 is Leu-1830, and some local difference in conformation probably exists. Like L1825P, the R1826H mutation leads to persistent current in Na V 1.5, further suggesting that the helix II-III interhelical segment is critical to channel inactivation.
Two mutations implicated in interactions with other components of the sodium channel cluster in helices III and IV. The D1866Y mutation in Na V 1.1, associated with generalized epilepsy and febrile seizures plus, leads to persistent current and decreased fast inactivation kinetics in the presence of the ␤ subunit (78). The corresponding position Asp-1856 in Na V 1.2 is at the start of helix IV and may disturb a putative surface for interaction with the ␤ subunit, as interaction with the ␤1 subunit and the CTD is suggested to occur through the second helix-sheet-helix motif by yeast-two-hybrid analysis of residues Lys-1846 -Arg-1886 in Na V 1.1 (78). Additionally, the M1852T mutation in Na V 1.1, also associated with generalized epilepsy and febrile seizures plus, results in decreased current (loss of function). This phenotype can be rescued by co-expression with ␤ subunits or calmodulin (79). Proposed to be a folding/ trafficking defect, this mutation may destabilize helix III, further suggesting that the second helix-sheet-helix motif may be important for interaction with other components of the sodium channel.
The notable exception to the above patterns is the LQT3 mutation D1790G in Na V 1.5, resulting in a relative negative shift in the voltage dependence of inactivation in the presence of the ␤ subunit (19,80). D1790G corresponds to position D1794 in helix I of Nav1.2 and may disrupt the helix by introduction of a glycine residue, with the effect of propagating to helices III and IV.
The mechanisms and extent of Na V 1 CTD function in binding the IQ motif and the specific role of calmodulin as well as Ca 2ϩ in multiple phases of inactivation remains to be elaborated. Interactions with the IQ motif may be more complicated than present models and may involve additional components (33). Previous evidence shows that Ca 2ϩ -dependent regulation of VGSC is mediated by calmodulin (31), with the exact mode of interaction yet to be determined. The solution structure of the Na V 1.2 C-terminal domain and chemical shift assignments of Na V 1.5 (1773-1878) are initial steps in elucidating the mechanism of inactivation, extended to other isoforms by virtue of high degrees of homology. The current work provides a tem-plate to begin probing specific interactions between the C-terminal domain and other components that play a role in inactivation of voltage-gated sodium channels.