Identification of the Cysteine Residue Responsible for Disulfide Linkage of Na+ Channel α and β2 Subunits*

Background: Voltage-gated Na+ channels are composed of α and β subunits. Results: We identified the cysteine residue in β2 responsible for disulfide linkage to α. Conclusion: α and β2 associate through a single disulfide bridge to achieve proper subcellular targeting in neurons. Significance: Understanding how Na+ channel complexes are formed in neurons is crucial for understanding the development of excitability. Voltage-gated Na+ channels in the brain are composed of a single pore-forming α subunit, one non-covalently linked β subunit (β1 or β3), and one disulfide-linked β subunit (β2 or β4). The final step in Na+ channel biosynthesis in central neurons is concomitant α-β2 disulfide linkage and insertion into the plasma membrane. Consistent with this, Scn2b (encoding β2) null mice have reduced Na+ channel cell surface expression in neurons, and action potential conduction is compromised. Here we generated a series of mutant β2 cDNA constructs to investigate the cysteine residue(s) responsible for α-β2 subunit covalent linkage. We demonstrate that a single cysteine-to-alanine substitution at extracellular residue Cys-26, located within the immunoglobulin (Ig) domain, abolishes the covalent linkage between α and β2 subunits. Loss of α-β2 covalent complex formation disrupts the targeting of β2 to nodes of Ranvier in a myelinating co-culture system and to the axon initial segment in primary hippocampal neurons, suggesting that linkage with α is required for normal β2 subcellular localization in vivo. WT β2 subunits are resistant to live cell Triton X-100 detergent extraction from the hippocampal axon initial segment, whereas mutant β2 subunits, which cannot form disulfide bonds with α, are removed by detergent. Taken together, our results demonstrate that α-β2 covalent association via a single, extracellular disulfide bond is required for β2 targeting to specialized neuronal subcellular domains and for β2 association with the neuronal cytoskeleton within those domains.

Voltage-gated Na ϩ channels are heterotrimeric complexes containing a single pore-forming ␣ subunit, a non-covalently linked ␤1 or ␤3 subunit, and a covalently linked ␤2 or ␤4 subunit (1). Although Na ϩ channel ␤ subunits are non-pore-forming, they play essential roles in channel targeting and regulation of channel gating and voltage dependence, and they play channel-independent roles in cell adhesion (2). Na ϩ channel ␤2 subunits (encoded by Scn2b) are disulfide-linked to the ion-conducting ␣ subunit and are expressed in brain, peripheral nerve, and heart (3)(4)(5). Similar to the other Na ϩ channel ␤ subunit family members, ␤2 is a type 1 transmembrane protein containing an extracellular N-terminal V-set Ig domain, a single transmembrane segment, and an intracellular C-terminal domain. In Xenopus oocytes, ␤2 modulates Na v 1.2 channel gating and increases functional channel expression through promotion of intracellular vesicle fusion with the plasma membrane as measured by changes in cell capacitance (3). In transfected fibroblasts, ␤2 increases Na v 1.2 cell surface expression only if ␤1 is also expressed (6). ␤2 functions as a trans homophilic cell adhesion molecule in vitro, resulting in recruitment of ankyrin to points of cell-cell contact (7). The extracellular ␤2 Ig domain associates with the extracellular domain of ␤1 in an in vitro cell adhesion assay, suggesting that these two subunits may associate via cell-adhesive interactions in vivo (8). Similar to the other Na ϩ channel ␤ subunits, ␤2 can be detected in detergent-resistant membrane preparations from primary neurons and is sequentially cleaved by ␤ and ␥ secretases (9). The cleaved intracellular domain of ␤2 is postulated to translocate to the nucleus to modulate Scn1a ␣ subunit gene transcription (10); however, the regulation of this process is not understood. A series of studies has demonstrated that ␤2 plays important roles in vivo in both neurons and cardiac myocytes. Scn2b null mice have significantly reduced (ϳ50 -60%) tetrodotoxin-sensitive Na ϩ current density in brain and small dorsal root ganglion (DRG) 5 neurons, altered voltage dependence of Na ϩ current inactivation, reduced sensitivity to pain, and increased suscep-tibility to seizures (5,11). Scn2b has been implicated in disease; Scn2b expression is up-regulated in a neuropathic pain model (12), and Scn2b null mice have a neuroprotective phenotype in a model of demyelinating disease (13). The Scn2b null mutation has been proposed to prevent pathologic Na ϩ channel up-regulation along axons in response to demyelination, thus attenuating the extent of neuronal degeneration. Although SCN2B has not yet been linked to human brain disease, mutations in SCN2B are associated with atrial fibrillation in the heart (14).
Concomitant covalent linkage of ␣ to ␤2 through disulfide bonding and insertion into the plasma membrane are the final steps in Na ϩ channel biosynthesis in primary neurons (15). These results suggested that covalent linkage of ␣ to ␤2 may occur at the extracellular face of the plasma membrane; however, the specific cysteine residue(s) involved have not been identified. Here we show, using a mutagenesis study, that a single cysteine-to-alanine substitution at the ␤2 extracellular residue Cys-26 is sufficient to disrupt the formation of ␣-␤2 covalent complexes in vitro. Although ␤2WT subunits do not affect the level of Na ϩ current expressed by Na v 1.1 in a heterologous system, co-expression of Na v 1.1 with ␤2C26A results in decreased Na ϩ current compared with ␣ alone, suggesting that ␤2C26A may cause intracellular retention of a population of ␣ subunits. Using a primary myelinating co-culture system, we demonstrate that, whereas ␤2WT traffics to nodes of Ranvier and heminodes, ␤2C26A is targeted to the axonal compartment but is not detectable at nodes or heminodes. In cultured hippocampal neurons, ␤2WT is enriched in the AIS, as defined by anti-ankyrin G staining, whereas ␤2C26A is expressed in a non-polarized distribution in all of the neuronal processes. Thus, covalent linkage of ␤2 to ␣ is essential for proper targeting of this subunit to specialized subcellular neuronal compartments. Finally, ␤2WT subunits are resistant to live cell detergent extraction from the hippocampal AIS, whereas mutant ␤2 subunits, which cannot form disulfide bonds with ␣, are removed by detergent. Taken together, our results demonstrate that ␣-␤2 covalent association via a single, extracellular disulfide bond is required for ␤2 targeting to specialized neuronal subcellular domains and for ␤2 association with the neuronal cytoskeleton within those domains.
Plasmids and Cell Culture-To generate a C-terminal ␤2V5 epitope-tagged expression plasmid, cDNA encoding rat ␤2 (minus the termination codon) was cloned into the multiple cloning site of pcDNA3.1/V5-His (ϩ) using standard TA cloning. This plasmid is referred to here as ␤2V5. Mutagenesis of ␤2 cysteine residues to alanine within this vector was performed using the QuikChange II site-directed mutagenesis kit (Stratagene). A subset of ␤2 constructs, including ␤2WT, ⌽ (the construct with all cysteine residues mutated to alanine), and ␤2C26A, were subcloned into pEGFP-N1 to add a C-terminal GFP epitope tag. The integrity of all plasmids was confirmed by DNA sequencing at the University of Michigan DNA Sequencing Core. HEK-293 cells stably expressing human Na v 1.1 (Gen-Bank TM accession number NP_008851.3; HEKhNa v 1.1) were obtained from GlaxoSmithKline under a materials transfer agreement as described previously (16).
Co-immunoprecipitation-Co-immunoprecipitation was performed similarly as described in Ref. 17. Briefly, HEKh-Na v 1.1 cells cultured as described (16) were transfected with 4 g of ␤2WT or mutant plasmid using Fugene 6 as recommended by the manufacturer. Protein A-Sepharose beads (Sigma) were washed with PBS and resuspended in 500 l of dilution buffer (DB) (60 mM Tris/HCl, pH 7.5, 180 mM NaCl, 1.25% Triton X-100, 6 mM EDTA, pH 8, containing Complete Mini protease inhibitor tablets (Roche Applied Science) at 2 times the manufacturer's recommended concentration) at 4°C. The beads were then incubated overnight at 4°C with 5 g of anti-pan-Na ϩ channel antibody. The transfected cells were detached from the culture dishes using 50 mM Tris, 10 mM EGTA (pH 8) and centrifuged at 5,000 rpm in a microcentrifuge for 5 min at 4°C. The cell pellet was resuspended in DB for cell lysis. After 30 min of lysis on ice, a 10-min centrifugation at 10,000 rpm in a microcentrifuge was performed to remove insoluble material. The resulting supernatant was added to the beads and incubated by rotating end-over-end for 5 h at 4°C. The beads were then washed twice with washing buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.1% Triton X-100, 0.02% SDS, 5 mM EDTA, pH 8, containing Complete Mini protease inhibitor tablets at 2 times the manufacturer's recommended concentration) followed by one wash with the same buffer lacking Triton X-100. Samples were then separated by SDS-PAGE on a 5% polyacrylamide gel and transferred to nitrocellulose for Western blot analysis. Western blotting was performed using the SnapId system (Millipore). Immunoblots were probed with anti-V5 or anti-pan-VGSC antibody as indicated, detected with Westfemto Chemiluminescent reagent (Pierce), and imaged using autoradiography film (Denville Scientific) or using the LI-COR Odyssey Fc imaging system. All results presented are representative of at least two independent repeats, as specified in the figure legends.
Surface Biotinylation-Surface biotinylation of ␤2WT or mutant proteins was performed as described previously (16). Briefly, membrane proteins were biotinylated using the Cell Surface Labeling Accessory Pack (Pierce) following the manufacturer's instructions utilizing Complete Mini (Roche Applied Science) as protease inhibitor. Samples were separated on 12% SDS-polyacrylamide gels. Proteins were transferred to nitrocellulose membranes that were processed for Western blotting using the SnapId system (Millipore). Anti-V5 immunoreactive signals detecting ␤2 subunits were compared with the immunoreactive signal for Na ϩ /K ϩ ATPase ␤1 subunit, a loading control for cell surface protein preparations (16). All results presented are representative of at least two independent repeats, as specified in the figure legends.
Immunocytochemical Analysis of ␤2 Expression in HEK Cells-HEKhNa v 1.1 cells were transiently transfected with GFP alone, ␤2WT-GFP, ␤2C26A-GFP, or ⌽-GFP using Fugene 6 following the manufacturer's recommendations. 24 h post-transfection, cells were replated onto 8-well glass chamber slides (BD Falcon). 24 h later, cells were fixed for 20 min at room temperature with 4% paraformaldehyde. After a 1-h blocking step (PBSTGS; dPBS containing 10% goat serum and 0.3% Triton X-100), cells were incubated overnight at room temperature with rabbit anti-GFP antibody (Invitrogen; A6455; 1:1000) diluted in PBSTGS. Following three 10-min washes with dPBS, cells were incubated for 2 h at room temperature with Alexa Fluor 568 goat anti-rabbit secondary antibody (Invitrogen; A-11011; 1:500) diluted in PBSTGS. After three 10-min washes with dPBS, slides were allowed to dry for 20 min and coverslipped using mounting medium (Invitrogen; ProLong Gold Antifade Reagent with DAPI; P36931). Slides were imaged using a Nikon A1R confocal microscope utilizing Nikon NIS-Elements software located in the Department of Pharmacology at the University of Michigan.
Whole-cell Patch Clamp Recording and Analysis-HEKh-Na v 1.1 cells were transfected with V5-or GFP-tagged ␤2 subunits as described above and plated for electrophysiological analysis. To detect cells that were transfected with V5-tagged plasmids, a 1:10 ratio of GFP/␤2 cDNA was used such that ␤2-expressing cells could be detected by epifluorescence. Aliquots of each transfection were analyzed for protein expression by Western blot to confirm protein expression. Micropipettes were obtained from capillary glass tubing (Warner Instruments) using a horizontal P-97 puller (Sutter Instruments). Micropipette resistance was between 1.5 and 3.5 megaohms when filled with intracellular solution containing 10 mM NaCl, 10 mM CsCl, 105 mM (cesium) aspartate, 10 mM EGTA, and 10 mM HEPES, pH 7.4, with CsOH and extracellular solution containing 130 mM NaCl, 4 mM KCl, 1.5 mM CaCl 2 , 1 mM MgCl 2 , 5 mM glucose, 10 mM HEPES, pH 7.4, with NaOH. Voltage pulses were applied, and data were recorded using Clampex version 9.2 or 10, an Axopatch 200B or 700B amplifier, and a Digidata 1322A digitizer (Molecular Devices). Pipette and whole cell capacitance were fully compensated. When appropriated, series resistance compensation was set to ϳ40 -70% with the lag set to 10 s. Signals were low pass-filtered at 5 kHz, and data were sampled at 40 kHz online. To determine the Na ϩ current amplitude and voltage dependence of activation, Na ϩ currents were evoked by 250 ms depolarizing test pulses (from Ϫ110 to 40 mV at 5 and 10 mV intervals) from a holding potential of Ϫ80 mV and a hyperpolarizing Ϫ120 mV, 250 ms prepulse. Peak Na ϩ current was normalized to cell capacitance and used to plot I-V curves and also to calculate conductance (g ϭ I/(V Ϫ V rev ), where V is the test potential, and V rev is the measured reversal potential). Voltage dependence of inactivation was determined by applying a 50 ms test pulse to 0 mV after 250 ms prepulses to the same voltages as described for the voltage dependence of activation. Peak currents were normalized to the maximum peak I Na amplitude. Normalized activation and inactivation curves were fit with a Boltzmann equation, 1/(1 ϩ exp(V Ϫ V1 ⁄ 2 )/), where V1 ⁄ 2 is the membrane potential in the midpoint of the curve, and as the slope factor. The kinetics of inactivation were measured during the test pulse to 0 mV from the same protocol used for voltage dependence of activation. The current, from 90% of the peak amplitude to 20 ms after the test pulse was initiated, was fitted to a double exponential equation of the form, I ϭ (F f ϫ exp(Ϫ/ f )) ϩ (Fs ϫ exp(Ϫ/ s ) ϩ C, using the Chebyshev method, where I is the current, f and s are the time constants for the fast and slow inactivation component, and C is the persistent current. Analysis of the recorded Na ϩ current was performed using the software packages Clampfit version 10 (Molecular Devices) and SigmaPlot version 11.2 (Systat Software, Inc., San Jose, CA). The statistical significance of differences between mean values for ␤2C26A-GFP or ⌽-GFP compared with ␤2WT-GFP was evaluated using Student's unpaired t test, with p Ͻ 0.05 considered significant. Results are presented as means Ϯ S.E.
Myelinating Co-cultures-Co-cultures of rat Schwann cells and DRG neurons were established as described previously (18). Briefly, DRGs were removed from Glu-15 or Glu-16 rat spinal cords, dissociated with 0.25% trypsin, and plated onto Matrigel (BD Biosciences)-coated coverslips. After cycling with antimitotics to eliminate non-neuronal cells, Schwann cells were added to the cultures and maintained in C medium (containing 10% fetal bovine serum, 50 ng/ml 2.5S nerve growth factor, 0.4% glucose, and 2 mM L-glutamine in minimum essential medium) for 1-3 days before adding 50 g/ml ascorbic acid to allow myelination to ensue.
Primary Hippocampal Neuron Cultures-Primary cultures of hippocampal neurons were established essentially as described previously (19). Briefly, hippocampi from embryonic day 18 rats were treated with 0.05% trypsin (Invitrogen) in dissecting solution (0.6% glucose, 10 mM HEPES in PBS) for 30 min at 37°C, and cells were dissociated by repeated passage through a fire-polished constricted Pasteur pipette and then plated onto 12-mm coverslips coated with poly-L-lysine (0.1 mg/ml in PBS) in minimal essential medium containing Earle's salts and glutamine with 10% FBS, 0.45% glucose, 1 mM pyruvate, and penicillin and streptomycin. After 2 h, the medium was replaced by Neurobasal medium (Invitrogen) with 2% B-27, and 0.5 mM L-glutamine. Cultures were maintained at 37°C in a humidified 5% CO 2 atmosphere until used.
Nucleofection-Nucleofector TM technology was used to introduce ␤2-GFP constructs into DRG neurons prior to coculture with Schwann cells or into hippocampal neurons. Nucleofection was performed with the Amaxa Rat Neuron Nucleofector Kit (Lonza) by using Nucleofector II (Lonza).
Immunofluorescence and Imaging-Hippocampal neurons or myelinating co-cultures were fixed in 4% paraformaldehyde for 10 min, permeabilized, and blocked with buffer containing 1% donkey serum, 5% BSA, and 0.2% Triton X-100 in 1ϫ PBS for 30 min and then stained with primary and secondary anti-Na ؉ Channel ␣-␤2 Disulfide Linkage NOVEMBER 9, 2012 • VOLUME 287 • NUMBER 46 bodies diluted in blocking buffer. Immunofluorescence images of cultures were taken with a Zeiss LSM 510 Meta confocal microscope.
Multiple Alignments of Na ϩ Channel Sequences-Multiple alignments of Na ϩ channel polypeptide sequences were performed using MEGA version 5 (21) utilizing the ClustalW methodology (22). NCBI GenBank TM accession numbers used in the alignment are provided in supplemental Table 1.

RESULTS
Na ϩ channel ␤2 subunits contain five extracellular cysteine residues (Cys-21, Cys-26, Cys-43, Cys-46, and Cys-98) and one intracellular cysteine residue (Cys-143) (3). Two of the extracellular cysteine residues form an intramolecular disulfide bridge as part of the Ig-fold structure (3). Aligning the amino acid sequence of ␤2 with ␤1 and ␤4, we found that ␤2Cys-21 and ␤2Cys-98 were conserved among the three ␤ subunits, suggesting that these residues are responsible for formation of the Ig loop intramolecular disulfide bridge that is common to all three (Fig. 1A). ␤3 was not included in this alignment because evidence suggests that it does not participate in homophilic cell-cell adhesion and thus may have a slightly different Ig domain structure (23). ␤2Cys-98 aligns with ␤1Cys-121, which is proposed to be critical to formation of the disulfide bridge in the Ig loop domain. ␤1Cys-121 is mutated to tryptophan in human epilepsy, resulting in disruption of ␤1-mediated cell-cell adhesion (24). These data support the idea that ␤2Cys-21 and ␤2Cys-98 participate in intramolecular disulfide bond formation in the Ig domain. Comparing ␤2 with ␤4 (25), the ␤ subunits that are disulfide-linked to ␣, we found that ␤2Cys-26 is conserved between subunits, whereas ␤2Cys-43 and ␤2Cys-46 are not conserved, suggesting that Cys-26 may be involved in disulfide linkage to ␣. In previous studies, we used the crystal structure of the myelin P 0 extracellular domain to model the Na ϩ channel ␤1 Ig loop (26). Employing a similar strategy here for the related ␤2 extracellular domain, we show that ␤2Cys-21 and ␤2Cys-98 are optimally aligned to form an intramolecular disulfide bond (Fig. 1B, left and right). In contrast, ␤2Cys-26 is located in the linker region between the B and C faces of the Ig loop (26), in a position that is more accessible for association with Na ϩ channel ␣ subunits (Fig. 1B, left). Thus, we hypothesized that ␤2Cys-26 was the most logical candidate to form an intermolecular disulfide bridge between ␤2 and ␣.
␤2Cys-26 Mediates the Disulfide Linkage between ␤2 and ␣-To identify the ␤2 cysteine residue(s) responsible for ␣-␤2 linkage, ␤2 mutant expression constructs were generated, substituting each single cysteine to alanine. In addition, every possible combination of multiple cysteine to alanine substitutions was generated, in the event that multiple cysteine residues were involved in ␣-␤2 linkage. Each construct was engineered to contain a C-terminal V5 epitope tag to facilitate biochemical identification. ␤2WT or cysteine-to-alanine ␤2 mutant constructs were then expressed separately in HEKhNa v 1.1 cells. Na v 1.1 was immunoprecipitated with anti-pan-Na ϩ channel antibody, and ␣-␤2 protein complexes (M r Ͼ 250,000) were detected by non-reducing SDS-PAGE followed by immunoblotting with a mouse monoclonal anti-V5 antibody to label V5-tagged ␤2 subunits. As expected (15), co-expression of ␤2WT with Na v 1.1 resulted in detection of a high M r ␣-␤2 channel complex (Fig. 2, A and B). Mutation of all six cysteine residues within ␤2 (⌽) completely disrupted ␣-␤2 association ( Fig. 2A). A single alanine substitution, C26A, was sufficient to disrupt the ␣-␤2 linkage ( Fig. 2A). Of all of the possible ␤2 cysteine-to-alanine mutant constructs, only those with the ␤2C26A mutation abolished ␤2-Na v 1.1 association under nonreducing conditions (Fig. 2B and Table 1). Importantly, mutation of ␤2Cys-21 or ␤2Cys-98, the residues proposed to form the intramolecular Ig loop disulfide bond, to alanine did not disrupt intermolecular ␣-␤2 subunit association. All of the cysteine-to-alanine mutant ␤2 constructs displayed robust expression levels in cell lysates collected post-transfection (␤2WT, ⌽, and ␤2C26A shown in Fig. 2C; others not shown). These data suggest that a single extracellular cysteine residue, ␤2Cys-26, mediates the disulfide linkage between ␤2 and ␣.
Cell Surface Expression and Modulation of Na ϩ Currents-The ␤2 mutant polypeptides ⌽ and ␤2C26A do not form disulfide bonds with Na v 1.1. Nevertheless, these subunits traffic to the cell surface, where they may be available to modulate Na ϩ currents through non-covalent association with ␣. Surface biotinylation analyses, shown in Fig. 3, A and B, demonstrated that ␤2WT, ⌽, and ␤2C26A are expressed at the cell surface. Fluorescence immunocytochemical data (Fig. 3C) confirmed cell surface expression of all three subunits. To test for the pos- sibility of non-covalent association of these mutant subunits with ␣, we co-immunoprecipitated HEKhNa v 1.1 cells transfected with ⌽ or ␤2C26A with anti-pan-Na ϩ channel antibody (as in Fig. 2) and probed the Western blot for non-disulfidelinked ␤2 subunits (migrating in the 33 kDa range) but were unable to detect any bands (not shown). Despite these results, we could not rule out the possibility that non-disulfide-linked ␤2 subunits, especially ␤2C26A, in which the Ig loop domain is predicted to remain intact, may associate transiently with ␣ subunits. To test this possibility, we investigated the effects of ␤2WT, ␤2C26A, or ⌽ on Na ϩ currents in transfected HEKh-Na v 1.1 cells under whole cell voltage clamp ( Table 2). The V5and GFP-tagged plasmids for each construct yielded similar results, and thus values were pooled. Similar to previous work with other Na ϩ channel ␣ subunits co-expressed with ␤2 in heterologous cells (6,14), we observed that co-expression of ␤2WT with Na v 1.1 did not change the level of Na ϩ current density compared with that observed for ␣ alone. Co-expression of ⌽ with Na v 1.1 also had no effect on Na ϩ current density compared with ␣ alone (Fig. 4). In contrast, we observed that co-expression of ␤2C26A with Na v 1.1 resulted in a significant decrease in transient Na ϩ current compared with WT (Fig. 4). Interestingly, these data are analogous to results reported for the SCN2B atrial fibrillation mutations R28Q and R28W, located in a similar region of the ␤2 Ig domain, which decreased the amplitude of Na v 1.5-expressed currents compared with ␤2WT (14). Also similar to (14) and to our results with Na v 1.5 expressed in a heterologous system (4), ␤2WT did not increase Na v 1.5-generated currents compared with ␣ alone (14). In addition to the effect of ␤2C26A on current density, we observed a small but significant depolarizing shift in the voltage dependence of Na ϩ current inactivation in the presence of ⌽ compared with ␣ alone. Thus, although we were unable to detect non-covalent association of ␤2C26A or ⌽ with Na v 1.1 biochemically, our electrophysiological results suggest that these subunits may associate with ␣ through low affinity interactions.
Disulfide Linkage of ␤2 with ␣ Is Critical for ␤2 Targeting to Nodes of Ranvier and the AIS-Na ϩ channel clustering at nodes of Ranvier is critical for action potential conduction in myelinated axons. Localization of ␤2 to nodes of Ranvier (28) suggests a functional role for this non-pore-forming subunit in these highly organized structures. Na ϩ channel ␣ subunits traffic normally to nodes of Ranvier in Scn2b null mice, demonstrating that ␤2 is not necessary for proper nodal localization and clustering of the ion channel pore. However, action potential conduction is significantly decreased in Scn2b null optic nerve, indicating that the density of Na ϩ channel ␣ subunits at the node is reduced compared with WT.
To test whether ␣-␤2 subunit disulfide linkage affects ␤2 targeting to nodes, we generated myelinating co-cultures consisting of DRG neurons and Schwann cells. Prior to co-culture, neurons were transfected with ␤2WT-GFP, ␤2C26A-GFP, or ⌽-GFP. C-terminal GFP epitope tags were added to the constructs to facilitate imaging of ␤2 localization. We examined the targeting of these constructs to heminodes (i.e. the initial nodal clusters that form at the ends of individual myelin segments) and to mature nodes, which are flanked on both sides by myelin segments. ␤2WT-GFP was targeted specifically to nodes of Ranvier as well as to the heminodes (Fig. 5A, top panels). Thus, all nodes (20 of 20) and all heminodes (44 of 44) scored in a representative experiment were labeled by the ␤2WT-GFP construct. These results also indicate that the C-terminal GFP tag does not interfere with normal ␤2 trafficking in neurons. In  Table 1. C, Western blot analysis shows robust expression of ␤2WT and mutant constructs. Top, HEKh-Na v 1.1 cells were transfected with V5-tagged ␤2WT, ⌽, or ␤2C26A or with GFP alone, as indicated. Cells were then solubilized and analyzed by Western blot with anti-V5 antibody to confirm subunit expression. Bottom, the blot was stripped and reprobed with anti-␣-tubulin as a loading control. Results presented in this figure are representative of five independent experiments.

TABLE 1 Detection of ␤2-Na v 1.1 disulfide linkage by coimmunoprecipitation
Coimmunoprecipitation was performed with HEKhNa v 1.1 cells transiently transfected with V5-tagged ␤2WT or various ␤2 cysteine-to-alanine mutants, as indicated. Solubilized Na v 1.1 complexes were immunoprecipitated with anti-pan-Na ϩ channel antibody and separated on non-reducing SDS-polyacrylamide gels as described under "Experimental Procedures." Western blots were detected with anti-V5 antibody to visualize high ␣-␤2 protein complexes, as demonstrated in Fig.  2. Results presented in this table are representative of two independent experiments, except for ␤2WT , ␤2C26A, ␤2C43A, ␤2C43A/C46A/C143A, ␤2C26A/C43A, and ␤2C26A/C43A/C46A/C143A, which were performed five times each. contrast to ␤2WT-GFP, ⌽-GFP (Fig. 5A, middle panels) and ␤2C26A-GFP (Fig. 5A, bottom panels) were diffusely expressed in the neurites of the sensory neurons but absent from the nodes and heminodes. In the case of ⌽-GFP, 0 of 52 heminodes and 0 of 21 nodes were positive; ␤2C26A-GFP was present in 0 of 48 heminodes and 0 of 12 nodes. Thus, although ␤2 subunits that are not covalently linked to ␣ still traffic to the neurites, they fail to cluster at nodes or heminodes.

Detection of immunoreactive
Na ϩ channel clustering at the AIS is critical for action potential initiation. We tested the targeting of ␤2WT-GFP, ⌽-GFP, and ␤2C26A-GFP to the AIS using primary hippocampal neuronal cultures. Similar to our results in myelinating co-cultures, we found that disruption of ␣-␤2 covalent linkage disrupted ␤2 targeting. ␤2WT-GFP was enriched at the AIS, defined by ankyrin-G expression, in 59 of 60 neurons scored. (It was exclusively expressed in the AIS in 5 of 60 neurons and faintly expressed in the remainder of the axon in 22 of 60 neurons and in all processes in 32 of 60 neurons) (Fig. 5B, top panels). In contrast, the ⌽-GFP (Fig. 5B, middle panels) and ␤2C26A-GFP (Fig. 5B, bottom panels) constructs failed to concentrate at the AIS in any neurons scored (i.e. 0 of 54 and 60 neurons, respectively). Rather, expression of both constructs was either nonpolarized or confined to the dendrites. Finally, we used the live cell detergent extraction technique (18) to investigate the association of ␤2 subunits with cytoskeletal elements at the AIS. As shown in Fig. 5C, ␤2WT subunits are resistant to Triton X-100 extraction from the hippocampal AIS (Fig. 5C, top panels). In contract ⌽-GFP (Fig. 5C, middle panels) and ␤2C26A-GFP   ( Fig. 5C, bottom panels), subunits that cannot form disulfide bonds with ␣ are removed completely by detergent treatment. Taken together, these results demonstrate that ␣-␤2 covalent association is required for ␤2 targeting to specialized neuronal subcellular domains and for ␤2 association with the neuronal cytoskeleton within those domains.

DISCUSSION
Voltage-gated Na ϩ channels are essential regulators of neuronal excitability in mammals (1). Concomitant covalent association of Na ϩ channel ␣ with ␤2 subunits and insertion into the plasma membrane are the final steps in channel biosynthesis in central neurons (15). Thus, ␣-␤2 association is considered to be a critical rate-limiting step in the formation of functional Na ϩ channels and consequently the development of excitability, playing an essential role in the regulation of channel density and channel targeting to specialized locations in the neuron (15). Studies with Scn2b null mice have demonstrated that ␤2 subunits are critical regulators of central and peripheral neuronal excitability in vivo via modulation of channel cell surface expression (5,11). Here we set out to close a critical gap in the Na ϩ channel literature: the molecular identification of the site of covalent ␣-␤2 association.
Via a combination of molecular biological, biochemical, and electrophysiological techniques, we show that the ␤2 residue Cys-26 is necessary and sufficient to mediate ␣-␤2 covalent association through disulfide bonding. We also expressed these constructs in primary hippocampal neurons and in myelinating co-cultures to investigate the role of covalent ␣ subunit association in ␤2 targeting to the AIS and nodes of Ranvier, respectively. Mutation of ␤2 extracellular residue Cys-26 to alanine resulted in disruption of ␣-␤2 covalent association. Despite the disruption of covalent ␣-␤2 association by this amino acid substitution, co-expression of ␤2C26A with Na v 1.1 decreased the level of transient Na ϩ current density compared with expression of ␣ alone, suggesting that ␤2C26A and ␣ may associate through transient, non-covalent interactions that we were unable to detect using biochemical techniques. The ␤2WT subunit is enriched in the AIS of hippocampal neurons and selectively expressed at nodes and heminodes in myelinating DRG-Schwann cell co-cultures. In contrast, ␤2C26A was not enriched in the AIS of hippocampal neurons but rather was diffusely expressed in both axons and dendrites or just in dendrites, indicating that its targeting was dramatically altered. Similarly, in myelinating co-cultures, ␤2C26A was never expressed at nodes but rather remained diffusely distributed along the neurites. Triton X-100 extraction of hippocampal neurons removed ␤2C26A from the AIS but left ␤2WT undisturbed, suggesting that ␤2 subunits are normally associated with the neuronal cytoskeleton and that disruption of ␣-␤2 covalent association eliminates ␤2 cytoskeletal interactions. These data suggest that ␣-␤2 subunit covalent association is essential for proper ␤2 clustering at specialized neuronal subcellular domains. In other words, ␤2 follows ␣. Further, ␤2 interactions with the cytoskeleton at the AIS may be mediated exclusively through Na ϩ channel ␣ subunits rather than through other cell adhesion molecules or via direct association with cytoskeletal proteins. We demonstrated previously, using a heterologous system, that ␤2 binds to the cytoskeletal protein ankyrin in response to cellular aggregation (7). Based on these results, we predict that ␤2-␤2 trans homophilic adhesion (e.g. through axonal fasciculation) may be required to transduce an outside-in signal to stimulate ␤2 association with ankyrin in neurons. In the absence of ␤2-␤2 trans adhesion, as is the case in the neuronal cultures used here, ␤2 does not associate with other ␤2 subunits on adjacent cells and thus does not bind ankyrin. In this case, ␤2 must depend on covalent linkage with ␣ for the cytoskeletal association. . Covalent ␣-␤2 linkage is critical for targeting of ␤2 to nodes of Ranvier and the AIS. A, targeting of ␤2 constructs to nodes of Ranvier. DRG neurons were nucleofected with WT or mutant ␤2-GFP constructs, as indicted, and then co-cultured with Schwann cells under myelinating conditions for ϳ2 weeks. Cultures were fixed and analyzed by immunofluorescence staining. ␤2WT-GFP (green) accumulated at nodes and heminodes (arrows), whereas the mutant constructs failed to accumulate at these sites. Paranodes were stained with Caspr (red), and myelin segments were stained with MBP (blue). Scale bar, 5 m. B, targeting of ␤2 constructs in hippocampal neurons. ␤2-GFP constructs were nucleofected into hippocampal neurons and analyzed at 18 days in vitro. ␤2WT-GFP (green) was enriched in the AIS, labeled by ankyrin G staining (red) (arrows). In contrast, ⌽-GFP and ␤2C26A-GFP mutants (green) were equally distributed in axons and dendrites or preferentially concentrated in dendrites and were not enriched in the AIS. Dendrites were stained with MAP2 (blue). Scale bar, 10 m. C, Triton X-100 extracts mutant ␤2 constructs from hippocampal neuron cultures. ␤2WT-GFP nucleofected hippocampal neuron cultures were extracted with Triton X-100 prior to fixation and then fixed and stained. The ␤2WT-GFP construct was retained at the AIS despite detergent treatment and was extracted from other sites. Both ⌽-GFP and ␤2C26A-GFP were largely extracted from the neurons, including from the AIS, by detergent treatment. Arrows in the GFP staining panels delineate the positions of AIS; neuronal somata are located on the left. Scale bar, 5 m.