Identification of a conserved ankyrin-binding motif in the family of sodium channel alpha subunits.

Interactions with ankyrinG are crucial to the localization of voltage-gated sodium channels (VGSCs) at the axon initial segment and for neurons to initiate action potentials. However, the molecular nature of these interactions remains unclear. Here we report that VGSC-alpha, but not -beta, subunits bind to ankyrinG using pull-down assays. Further dissection of this activity identifies a conserved 9-amino acid motif ((V/A)P(I/L)AXXE(S/D)D) required for ankyrinG binding. This motif is also required for the localization of chimeric neurofascin/sodium channel molecules to the initial segment of cultured hippocampal neurons. The conserved nature of this motif suggests that it functions to localize sodium channels to a variety of "excitable" membrane domains both inside and outside of the nervous system.

Interactions with ankyrin G are crucial to the localization of voltage-gated sodium channels (VGSCs) at the axon initial segment and for neurons to initiate action potentials. However, the molecular nature of these interactions remains unclear. Here we report that VGSC-␣, but not -␤, subunits bind to ankyrin G using pull-down assays. Further dissection of this activity identifies a conserved 9-amino acid motif ((V/A)P(I/L)AXXE(S/D)D) required for ankyrin G binding. This motif is also required for the localization of chimeric neurofascin/sodium channel molecules to the initial segment of cultured hippocampal neurons. The conserved nature of this motif suggests that it functions to localize sodium channels to a variety of "excitable" membrane domains both inside and outside of the nervous system.
The concentration of voltage-gated sodium channels (VGSCs) 1 into excitable membrane domains is crucial to information processing and transmission in the nervous system and to excitation/contraction coupling in muscle. Localized concentrations of VGSCs at the initial segment (IS) and the nodes of Ranvier are necessary for the initiation and propagation of action potentials through myelinated axons. However, mechanisms underlying the localization of VGSCs to these excitable membrane domains remain relatively unknown. In neurons, VGSCs exist as heterotrimers composed of a large pore-forming ␣ subunit associated with two smaller accessory ␤ subunits (reviewed in Ref. 1). At least 10 genes encoding putative ␣ subunits have been identified in mammals (reviewed in Ref. 2) along with three ␤ subunit genes. ␤ subunits not only modulate the activity of ␣ subunits (3,4) but also exhibit characteristics of cell adhesion molecules (CAMs) binding to extracellular matrix molecules such as tenascin-C and -R (5) and other CAMs such as neurofascin (6).
Polarized localization of VGSCs in the axonal membrane requires interactions with members of the ankyrin family of peripheral membrane proteins (reviewed in Ref. 7). Ankyrins function as membrane-cytoskeleton adaptors that immobilize integral membrane proteins to the spectrin-based membrane skeleton. Ankyrins co-purify with and directly bind to purified VGSCs (8), and specific isoforms of the ankyrin G gene are concentrated with VGSCs at the IS (9), the nodes of Ranvier (9), and the neuromuscular junction (10). The functional importance of the interaction between ankyrin G and VGSCs is demonstrated in knockouts of ankyrin G in the mouse cerebellum. Purkinje cells from knockout animals are deficient in localized VGSC concentrations at the IS (11) and are unable to initiate action potentials (12).
In addition to VGSCs, ankyrins also interact with a variety of other integral membrane proteins present at the node including the CAMs neurofascin and NrCAM (13,14). Interactions of ankyrins with VGSCs (15) and other integral membrane proteins are mediated through the N-terminal 90-kDa repeat or membrane-binding (MB) domain. The interaction between the ankyrin MB domain and neurofascin/NrCAM is dependent upon a conserved FIGQY motif in the cytoplasmic domain of these molecules (16).
Unlike neurofascin and NrCAM, the nature of the interaction between VGSCs and ankyrins remains unclear. Initial studies using uncharacterized rat brain preparations of VGSCs and purified erythrocyte ankyrin (15) did not distinguish which VGSC subunits or their isoforms were involved in ankyrin binding. Subsequent studies using a variety of techniques have suggested sites for ankyrin interaction on both the VGSC ␤ (17) and ␣ subunits (18).
In this study, we identified a domain of the VGSC-␣ subunit that interacts with ankyrin G and demonstrated its role in the localization of VGSCs to the IS in vivo. Initially we observed that rNa v 1.2 (rat Na v 1.2) co-localizes with ankyrin G at the IS of hippocampal neurons and identified an ankyrin G -binding site in the cytoplasmic linker between domains II and III (loop 2), using a pull-down assay. We did not detect binding sites for ankyrin G elsewhere within the ␣ subunit nor did we detect binding with the ␤ subunits using this assay. We further identified a 9-aa motif within the loop 2 sequence that is conserved among all vertebrate ␣ subunits and demonstrated its requirement in ankyrin G binding. Finally, we demonstrated that deletion of this 9-aa motif abolishes the ability of neurofascin/loop 2 chimeric molecules to be localized to the IS of cultured hippocampal neurons. These results demonstrate that ankyrin binding is conserved among vertebrate sodium channels and is crucial to the localization of VGSCs to excitable membrane domains. from the hippocampi of 18-day-old fetal Wistar rats and transfected using a modified calcium/phosphate protocol as described (19).
Construction of Plasmids for Protein Expression-Rat cDNAs for the ␤1 and ␤2 subunits of the voltage-gated sodium channels (a gift from Dr. L. L. Isom, University of Michigan) were subcloned into pEGFP-N1 (BD Biosciences) (not in frame with the downstream GFP), and a sequence coding for the HA epitope (YPYDVPDYA) was inserted by PCR at a position corresponding to aa 26 for ␤1 and 34 for ␤2 (see , respectively. A partial cDNA for rNa v 1.5 was obtained by reverse transcriptase-PCR using the Prostar kit (Stratagene, La Jolla, CA) on rat heart mRNA isolated with the Oligotex kit (Qiagen, Inc., Chatsworth, CA) following the manufacturers' specifications. Deletion of consensus sequence VPIAXX-ESD in constructs II-III and II-III (1.5) was obtained by "loop-out" mutagenesis using the QuikChange kit (Stratagene) following the manufacturer's instructions (constructs II-III⌬ and II-III⌬ (1.5), respectively; see Fig. 3).
Neurofascin/Na v Chimeric Molecules-Full-length cDNAs encoding HA-tagged neurofascin 186-kDa isoforms were a gift from Dr. Vann Bennett (Duke University, Durham, NC), and their preparation has already been described (20). The 500-bp ApaI-NotI fragment representing the cytoplasmic domain of neurofascin was subcloned into pBluescript (Stratagene), and the Y81A mutation was generated using the QuikChange mutagenesis kit (Stratagene) before the altered cytoplasmic domain fragments were re-introduced into full-length neurofascin molecules. PCR products representing the loop 2 sequence from Na v 1.2a and -1.2⌬ (Na v 1.2a missing the VPIAXXESD motif) were modified to include a 5Ј in-frame ApaI site and a 3Ј stop codon upstream of a NotI site. These products were then substituted for the neurofascin-cytoplasmic domain in full-length HA-tagged neurofascin (NF) 186-kDa molecules to generate neurofascin/Na v chimeric molecules (see Figs. 3

and 4).
Sindbis Pseudovirions-PCR was used to generate a 3Ј fusion of the 24-ankyrin repeats from rat ankyrin G (MB, aa 1-853) with sequenceencoding glutathione S-transferase (GST). The cDNA for MB-GST, or GST alone, was cloned into the pSinRep5 vector (Invitrogen) and used as a template to synthesize capped RNA transcripts in vitro with the Invitroscript CAP kit (Invitrogen). The transcripts were electroporated into BHK cells along with similarly prepared transcripts encoding the structural genes for the DH(26S) helper phage. BHK cells were then cultured in ␣-MEM with 5% fetal bovine serum for 48 h, and the media containing Sindbis pseudovirions for MB-GST or GST were collected, aliquoted, and kept at Ϫ80°C.
In Vitro Ankyrin G Binding Assay-Media enriched with Sindbis pseudovirions for MB-GST or GST were diluted in serum-free ␣-MEM, and optimal viral dilutions were determined to obtain high and reproducible levels of protein expression. 450 l of these dilutions were applied to Ϸ10 6 BHK cells in 60-mm dishes and incubated for 1 h at 4°C with intermittent rocking. Then 4 ml/dish ␣-MEM with 1% FBS (v/v) was added, and the cells were incubated for a further 24 h at 37°C. Cells were then washed twice in phosphate-buffered saline and incubated for 20 min on ice with 600 l/dish lysis buffer (50 mM Tris-Cl, 150 mM NaCl, 1% Triton X-100, 0.05% Tween 20, 1 mM Na 2 EDTA, pH 8.0) in the presence of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 g/ml each antipain, leupeptin, and pepstatin). Lysates were centrifuged, and the supernatant containing MB-GST or GST was incubated with one-fifth volume of glutathione-Sepharose 4B in a 50% slurry (Amersham Biosciences) for 30 min at 4°C on a rocker. The beads were then washed three times in lysis buffer and kept on ice before use.
For pull-down experiments with MB-GST or GST alone as a control, BHK cells were transfected with GFP-Na v or neurofascin/Na v chimeric constructs as described above and incubated for 24 h at 37°C. Cells were then lysed as above in 600 l/60-mm dish of lysis buffer with protease inhibitors. 300 l of cell lysate was then added to either 20 l of MB-GST beads or 20 l of GST beads and rocked overnight at 4°C. The concentration of MB-GST or GST used in each pull-down was typically 2 M. After incubation, the beads were washed three times with ice-cold lysis buffer, and bound proteins were eluted twice with 1 bed volume of 10 mM reduced glutathione in 50 mM Tris-Cl, pH 8.0. Cell lysates and proteins associated with MB-GST and GST were probed by immunoblotting using antibodies to GFP (Covance) or the HA tag (Covance). Similarly, cell lysates were analyzed to ensure expression of the tagged protein, and batches of beads used in the assays were analyzed using antibodies to GST to ensure the presence of the MB-GST and GST proteins (not shown).
Yeast Two-hybrid System-The MB domain from ankyrin G was subcloned in frame with the Gal4 activation domain into LEU2-marked prey vector pGAD424 (a gift from Dr. Peter Pryciak, University of Massachusetts Medical School, Worcester, MA). The loop 2 EcoRI cassettes described above were subcloned in frame with the LexA DNAbinding domain into TRP1-marked bait vector pBTM116. The cytoplasmic domain of neurofascin in its wild type form or carrying the Y81A mutation was also subcloned into pBTM116 to be used respectively as the positive or negative control for interaction with MB. Yeast strain L40, harboring the HIS3 reporter gene under the control of LexAbinding sites, was co-transformed with 50 ng each of the bait and prey vector using the lithium/acetate method and then streaked out onto agar plates lacking tryptophan and leucine. For each transformation, six independent clones were selected and replica-plated onto media lacking tryptophan, leucine, and histidine to assess expression of the HIS3 reporter gene.

Na v 1.2 Co-localizes with Ankyrin G in Vivo and Binds to Its
Repeat Domain in Vitro-Rat hippocampal neurons in culture (11 days in vitro) were double-labeled with an antibody against the conserved tail region of ankyrin G 480/270 kDa (Fig. 1a, ank G 480/270 kDa) and with a specific antibody against the C-terminal end of rNa v 1.2 (Fig. 1a, rNa v 1.2). Neurons exhibited localized concentrations of rNa v 1.2 at their IS coincident with similar concentrations of ankyrin G 480/270 kDa.
We developed a pull-down assay to determine ankyrin Gbinding sites within defined subunits and domains of rNa v 1.2a using a GST fusion of the repeat domain (MB) of ankyrin G (MB-GST). This construct was expressed at high levels in BHK cells using the Sindbis viral expression system and immobilized on glutathione-Sepharose beads. Similar beads coupled to GST alone were used as controls for each assay. To test for interactions with ␤ subunits, beads were added to cell lysates from BHK cells expressing HA-tagged ␤1 and ␤2 subunits (Fig.  1b). As shown in Fig. 1b, these subunits did not exhibit interactions with MB-GST.
To test for interactions with the rNa v 1.2a ␣ subunit, we incubated MB-GST and GST beads with BHK cell lysates expressing full-length rNa v 1.2a. However, we were unable to obtain efficient solubilization of the full-length rNa v 1.2a protein under the non-denaturing conditions used in our binding assay (data not shown). We therefore generated a series of GFP-tagged constructs that spanned rNa v 1.2a (Fig. 1c) in which we tried to maintain the structure of the cytoplasmic loops while preserving the integrity of the transmembrane domains (Fig. 1c). Constructs spanning the rNa v 1.2a ␣ subunits III-IV, II-III, and the III domain all precipitated with MB-GST beads and not with GST beads alone (Fig. 1c). This interaction did not appear to involve the C terminus of rNa v 1.2a as deletion of this region of the molecule from construct III-IV (construct III-IV⌬C) did not interfere with its ability to interact with MB-GST. In contrast, domain I-II of rNa v 1.2a and a GFP fusion of the cytoplasmic C terminus alone (not shown) did not exhibit interactions with MB-GST.
These results suggested the presence of a site of ankyrin G interaction between aa 1044 and 1474 of rNa v 1.2a. Further studies in which individual cytoplasmic loops of the III-IV construct were replaced sequentially with the nonbinding N terminus of rNa v 1.2a confirmed that the binding site was restricted to loop II-III (not shown). These results are in contrast to the findings of Bouzidi et al. (18) who recently reported interactions between ankyrin G and fusion proteins representing the isolated intracellular loops between domains I-II and III-IV but not between the II-III loop and ankyrin G .
Ankyrin G Associates with a Conserved 9-aa Motif in Loop 2 of VGSC-␣ Subunits-To further define the ankyrin G -binding site within loop 2, we used deletion analyses. An internal deletion of aa 1024 -1121 in loop 2 or its substitution with a portion of loop 1 (aa 599 -694) (Fig. 2a, constructs II-⌬-III and II-1-III, respectively) abolishes the interaction with MB-GST restricting the binding site to aa 1044 -1121 of rNa v 1.2a. Fig. 2c shows an alignment of the sequence of loop 2 between rat VGSCs. A high degree of homology is observed between aa 1024 -1121 of rNa v 1.2a and the analogous region of other neuronal isoforms (87, 76, 71, and 65% identity with rNa v 1.1, -1.3, -1.6, and -1.7, respectively) with the exception of rNa v 1.8 and rNa v 1.9 (only 15 and 26% identity, respectively). This region is less well conserved between rNa v 1.2a and the skeletal muscle and cardiac isoforms rNa v 1.4 (41% identity) and rNa v 1.5 (16% identity), respectively.
We decided to exploit the range of conservation within various loop 2 sequences to further map the site of ankyrin G interaction. We substituted aa 1024 -1121 in construct II-III with the analogous region (Fig. 2c) 5)). These constructs were expressed in BHK cells and used in pull-down assays with MB-GST as before. Chimeras II-III (1.1), II-III (1.6), and II-III (1.4) were specifically precipitated by MB (Fig. 2b). Despite its low level of homology, chimera II-III (1.5) also retained the ability to bind to MB (Fig. 2b). Careful analysis of the sequence of these five isoforms revealed only one stretch of nine amino acids conserved in this region, VPIAXXESD (indicated in bold in Fig. 2c), corresponding to aa 1105-1113 in rNa v 1.2a and located within the binding region defined previously.
We selectively deleted this sequence of 9 aa in constructs II-III and II-III (1.5) using loop-out mutagenesis to obtain the deletion mutants II-III⌬ and II-III⌬ (1.5), respectively. As shown in Fig. 2d, deletion of this 9-aa motif resulted in a loss of binding to MB-GST. The sequence VPIAXXESD is thus necessary for association of both II-III and II-III (1.5) with MB and defines a binding sequence for ankyrin G within the ␣ subunit rNa v 1.2a. Fig. 2e shows that this motif is conserved among rat VGSCs regardless of the overall degree of homology in loop 2. We have found that its presence is associated with binding to MB-GST in at least six isoforms (indicated with a "ϩ" in To test whether loop 2 and the conserved 9-aa motif were sufficient and necessary for ankyrin G association, we constructed chimeric molecules consisting of HA-tagged neurofascin extracellular domain fused to the loop 2 sequence (Fig. 3a).  Bound proteins (left) or 5% of the total lysate used for each pull-down (right) was analyzed by immunoblotting using an antibody to the HA epitope. c, top panel, diagrammatic representation of the VGSC-␣ subunit aligned above a linear representation of GFP-tagged constructs used in pull-down experiments. Numbers refer to the position of the corresponding amino acid residue in rNa v 1.2a. Bottom panel, lysates of cells expressing GFP-tagged constructs were incubated with MB-GST (ϩ lanes)-or GST (Ϫ lanes)conjugated beads as described for the ␤ subunits. Bound proteins (left) or 6% of the total lysate used in each pull-down (right) was analyzed by immunoblotting with a GFP antibody. Higher molecular weight denaturation-dependent aggregation artifacts of the expressed rNa v 1.2a subdomains were also observed. Similar artifacts have been reported previously for transiently expressed VGSC-␣ subunits in human embryonic kidney cells (34). lacking the 9-aa motif (rNa v 1.2⌬) were utilized (Fig. 3a). These chimeric constructs along with wild type neurofascin 186 kDa were used in pull-down assays with MB-GST. As a negative control, we mutated the tyrosine residue within the FIGQY motif of wild type neurofascin to an alanine (NF Y81A). Phosphorylation of this tyrosine residue has been previously shown to abolish ankyrin binding (20). Fig. 3a shows that chimeric molecules containing loop 2 sequences from rNa v 1.2a (NF 1.2) exhibited interactions with MB-GST suggesting that loop 2 alone was sufficient to promote association with ankyrin G . However, deletion of the conserved 9-aa motif (NF 1.2⌬) led to a loss of MB-GST interaction. As expected, wild type neurofascin associated strongly with MB-GST, whereas mutation of the tyrosine residue to alanine (NF Y81A) abolished this interaction despite comparable expression levels (Fig. 3a). The observation that a single aa change can abolish ankyrin binding illustrated the sensitivity and specificity of our pull-down assay using MB-GST. Our data demonstrate that loop 2 and the conserved 9-aa motif are necessary for ankyrin G binding.
To determine whether the loop 2/ankyrin G interaction involved proteins specific to BHK cell lysates, we carried out a yeast two-hybrid reaction. As shown in Fig. 3b, the loop 2 sequence from rNa v 1.5 fused to the LexA DNA-binding domain (pBTM Na v 1.5) was able to activate expression of the L40 HIS3 reporter gene when co-transfected with the MB sequence fused to the Gal4 activation domain (pGAD MB). His3 was not expressed when pBTM Na v 1.5 was co-transfected with the Gal4 activation domain alone nor when a mutated version of rNa v 1.5 lacking the 9-aa motif (pBTM Na v 1.5⌬) was used. The wild type neurofascin cytoplasmic domain and the Y81A variant of this domain were used as positive and negative controls, respectively (Fig. 3b). Similar studies involving loop 2 sequences from Na v 1.2 and -1.6 could not be carried out as these constructs were observed to transactivate reporter genes in the absence of pGAD MB.
The Conserved 9-aa Motif Affects Localization to the IS of Cultured Hippocampal Neurons-To determine whether loop 2 and the 9-aa motif affected localization of membrane proteins at the IS, we utilized the neurofascin/loop 2 chimeric constructs. These molecules were transfected into 3 days in vitro hippocampal cells, and these cells were fixed and stained with HA antibodies at various time points after transfection to determine when and where these proteins could be first detected on the cell surface. Cells were then permeabilized and stained with antibodies to ankyrin G 480/270 kDa to delineate the IS. As Bound proteins to MB-GST (ϩ lanes) or GST (Ϫ lanes) (left) or 6% of the total lysate used in each experiment (right) was analyzed by immunoblotting with a GFP antibody. c, the sequences from loop 2 of rNa v 1.1, rNa v 1.6, rNa v 1.4, and rNa v 1.5 that were used to generate the chimeras used in panel b are shown aligned with the sequence they replaced in construct II-III. Numbers correspond to the position of the first amino acid in each line for the corresponding isoform. Shaded residues are identical in at least three sequences. The motif VPIAXXESD is conserved (bold). d, the sequence VPIAXXESD was deleted from the II-III and II-III (1.5) constructs to give II-III⌬ and II-III⌬ (1.5), respectively. These constructs were subject to pull-down assays as before. Bound proteins (left panel) and 6% of total lysate used (right panel) were analyzed by immunoblotting with antibodies for GFP. e, conservation of the binding motif between all known rat voltage-gated sodium channels. Numbers correspond to the position of the first residue of the motif within each isoform. Shaded residues are identical in at least six sequences. Sequences tested and found positive for interactions with ankyrin G are indicated with a plus sign.
shown in Fig. 4, ϳ75% of neurons 6 h after transfection with neurofascin/loop 2 chimeric molecules (NF 1.2) exhibited concentrated HA staining at their IS (Fig. 4A) as delineated by ankyrin G staining (Fig. 4A, inset). Staining was also observed on the cell body of transfected neurons. 24 h after transfection, ϳ90% of the neurons now exhibited HA concentrations at the IS, and neuronal cell body staining was diminished compared with the signal at the IS (Fig. 4E). In contrast to wild type loop 2 molecules, chimeric molecules lacking the 9-aa motif (NF 1.2⌬) were observed over the surface of the entire neuron and did not exhibit concentration at the IS even at later time points (Fig. 4, B and F). Indeed, at earlier time points NF 1.2⌬ showed increased concentrations at the distal end of the axon (Fig. 4B).
We also introduced wild type and Y81A neurofascin constructs into hippocampal neurons. Unlike NF 1.2, wild type neurofascin did not immediately exhibit concentrations at the IS except in a small population of neurons (Ͻ10%). Instead, HA staining was distributed throughout the cell (Fig. 4G) in a similar fashion to that observed for the Y81A mutant (Fig. 4D). However, NF was observed to accumulate over time at the IS, such that between 45 and 65% of transfected neurons showed IS accumulations of neurofascin at 24 h after transfection (Fig.  4G). Similar concentrations were not observed for the Y81A mutation (Fig. 4H) supporting the idea that concentration of chimeric constructs at the IS is independent of the neurofascin extracellular domain. These results indicate that the 9-aa motif is necessary for the specific localization of neurofascin/loop 2 chimeric constructs at the IS. DISCUSSION In this study, we describe for the first time the identification of a conserved motif in the family of VGSC-␣ subunits neces-sary for their binding to ankyrin G . Initially, we observe that ankyrin G co-localizes with rNa v 1.2 at the IS of cultured hippocampal neurons and interacts with the domain II-III cytoplasmic linker (loop 2) of rNa v 1.2a. This represents the first specific function that has been ascribed to this area of VGSC-␣ subunits. We provide evidence to show that ankyrin G binding may be a common feature of the VGSC-␣ subunit family and further identifies a conserved 9-aa motif within loop 2 sequences that is necessary for ankyrin G binding. Finally, we demonstrate that the 9-aa motif is required for the localization of neurofascin/loop 2 chimeric molecules to the IS of cultured neurons. Based on these observations, we propose that the 9-aa motif functions in the concentration of VGSCs within excitable membrane domains through interactions with ankyrin.
Recent studies have yielded conflicting data on the nature of the interaction between ankyrin and VGSCs. Regulated interactions between ankyrin and VGSC-␤ subunits have been detected using a cell-based recruitment assay (17,21). In contrast to these results, Bouzidi et al. (18) did not detect interactions between ankyrin G and the cytoplasmic domains of the ␤ subunits using bacterially expressed proteins in an in vitro binding assay. However, they did detect interactions with recombinant fusion proteins representing the I-II (loop 1) and the III-IV (loop 3) cytoplasmic linkers. They detected no interactions with loop 2. In our study, we also did not detect interactions with the VGSC-␤ subunits using ankyrin G, but detected an interaction with the loop 2 cytoplasmic linker of the ␣ subunit. We further confirmed this interaction using yeast two-hybrid and demonstrated that loop 2 was sufficient for the co-localization of chimeric molecules at the IS with ankyrin G . Differences between our findings and those of Bouzidi et al. (18) might reflect their use of bacterially expressed polypeptides. In support of this idea, we have been unable to detect interactions between bacterially expressed ankyrin G and loop 2 in an in vitro assay.
To further examine the effects of loop 2 on VGSC-␣ subunit localization we have used chimeric neurofascin/Na v 1.2 molecules and observed that these molecules localize at the IS in the presence of the 9-aa motif. The observation that chimeric molecules are also initially detected on the neuronal cell body but are down-regulated there over time (Fig. 4) suggests that the retention of chimeric molecules at the IS is important for their concentration. However, ankyrin binding could also play a role in the trafficking of these chimeric molecules to the initial segment as has been suggested for the intracellular sorting of Ca 2ϩ -homeostasis proteins in cardiomyocytes (22).
The finding of a conserved motif in VGSCs involved in ankyrin G interaction is supported by increasing evidence for spatial and temporal diversity in VGSC isoforms enriched at ankyrin G -defined excitable membrane domains. Both Na v 1.1 and Na v 1.2 have been observed at the nodes of Ranvier in young mice and are gradually replaced with Na v 1.6 during development (23,24). Na v 1.6 is the VGSC isoform found at most adult nodes in the peripheral nervous system and the central nervous system (25); however, Na v 1.2, -1.8, and -1.9 have also been detected at adult nodes (24,26,27). A similar diversity is observed at the IS where Na v 1.2, -1.6, and -1.8 have been localized (11,26). In other tissues, ankyrin G is also localized to the neuromuscular junction (10) where it is a candidate to interact with skeletal isoforms of the sodium channel such as Na v 1.4. The observation that ankyrin G can also interact with cardiac Na v 1.5 suggests that isoforms of ankyrin G might also be linked with heart disease as has been shown for ankyrin B in Long QT syndrome (28).
What are the mechanisms responsible for the segregation of specific VGSC isoforms with ankyrin G in excitable membrane domains? The distinct spatial and temporal pattern of expres-FIG. 3. Loop 2 is sufficient and the 9-aa motif necessary to promote ankyrin G association. a, chimeric neurofascin/loop 2 molecules. Diagrammatic representation (top panel) of chimeric neurofascin molecules using loop 2 sequences from Na v 1.2 and Na v 1.2⌬ (a deletion of the VPIAXXESD motif). NF Y81A was used as a negative control for ankyrin G association. These constructs (bottom panel) were expressed in BHK cells and the lysates incubated with MB-GST (ϩ lanes)-or GST (Ϫ lanes)-conjugated beads. Bound proteins (left) or 6% of the total cell lysate used (right) was analyzed by immunoblotting with antibodies to HA. b, Saccharomyces cerevisiae L40 strain was co-transformed with pBTM116-Na v 1.5 or pBTM116-Na v 1.5⌬ (lacking the 9-aa motif) and pGAD424 alone or pGAD424-MB. The cytoplasmic domain of wild type neurofascin or the Y81A mutation was used as a positive or negative control, respectively, for the interaction with MB. Co-transformants were selected on solid media lacking tryptophan and leucine and then tested for activation of the HIS3 reporter gene. sion of VGSC isoforms in nervous tissue (29) is clearly a factor in determining the molecular identity of the VGSC present at the nodes of Ranvier and the IS. Indeed, the gradual isoform switch from Na v 1.2 to Na v 1.6 at developing nodes is paralleled by a down-regulation of Na v 1.2 and an up-regulation of Na v 1.6 proteins (24). The selective segregation of VGSCs with ankyrin G could also result from isoform-dependent differences in the affinities of interaction that are not detectable by the methods used in this study. In addition, post-translational modifications similar to those reported for neurofascin (20) could regulate the ankyrin/VGSC interaction and play a role in determining VGSC specificity.
Additional sequences within the ␣ subunit could also influence localization of specific VGSC isoforms within neurons. For example, the C terminus of rNa v 1.2 targets CD4 chimeric molecules to the axon of hippocampal neurons, whereas distribution of a similar construct derived from the C terminus of rNa v 1.6 is restricted to the somatodendritic region (30). The role of accessory subunits in VGSC localization should also be considered. ␤ subunits appear to influence the surface expression of VGSC isoforms (31) and are suspected to exhibit selective associations with different ␣ subunits (24). Similarly, the association of annexin II light chain with the N terminus of Na v 1.8 increases its translocation to the plasma membrane (32). Thus, accessory subunits may participate in the functional localization of VGSC isoforms by promoting their specific insertion into neuronal membranes.
Given the high degree of homology between the repeat domains of different ankyrin molecules, we might also expect VGSC-␣ subunits to interact with other ankyrins such as ankyrin B or ankyrin R . Indeed, ankyrin R has been shown to bind to uncharacterized native VGSC preparations from the rat brain (8). However, it is yet to be determined whether the affinities for a specific VGSC isoform vary between different ankyrins.
Data base searches with a consensus sequence for the 9-aa ankyrin-binding motif found this sequence to be present only in members of the VGSC-␣ subunit family, suggesting that this particular ankyrin-binding motif is unique to these proteins. This is in agreement with current models of ankyrin interactions proposing that different families of ankyrin-binding proteins independently evolve the ability to bind ankyrin (16). These searches also found the motif to be present in VGSC-␣ subunits from non-mammalian vertebrates such as the puffer fish, Takifugu pardalis, and the newt. Intriguingly, this motif was not found in the loop 2 ␣-subunit sequences of invertebrate sodium channels such as those observed in Drosophila melanogaster. It would therefore be of some interest to determine whether or not VGSC/ankyrin-enriched excitable domains can be detected in Drosophila.
Observations in knockout animals have demonstrated the role of the cytoskeletal elements ankyrin G and ␤IV spectrin in establishing and/or maintaining a functional axon IS required for initiation of an action potential (11,12,33). The 9-aa motif described in this study links VGSCs to these cytoskeletal elements and as such is necessary to confer the excitable dimension to this domain. The highly conserved nature of the motif suggests a complex regulation of VGSC/ankyrin association to achieve isoform specificity but provides a universal code for the establishment of VGSC-enriched membrane domains through ankyrin interaction that is likely to be common to a variety of tissues and organisms.