Specific Role of the Truncated βIV-Spectrin Σ6 in Sodium Channel Clustering at Axon Initial Segments and Nodes of Ranvier*

At axon initial segments and nodes of Ranvier in neurons, the spectrin membrane skeleton plays roles in physically stabilizing the plasma membrane integrity and in clustering voltage-gated sodium channels for proper conduction of the action potential. βIV-Spectrin, an essential component of the membrane skeleton at these sites, has an N-terminal-truncated isoform, Σ6, which is expressed at much higher levels than the full-length isoform Σ1. To investigate the role of βIV-spectrin Σ6, we generated Σ1-deficient mice with a normal level of Σ6 expression (Σ1-/- mice), and compared their phenotypes with those of previously generated mice lacking both Σ1 and Σ6(Σ1Σ6-/- mice). The gross neurological defects observed in Σ1Σ6-/- mice, such as hindleg contraction, were apparently ameliorated in Σ1-/- mice. At cellular levels, Σ1Σ6-/- and Σ1-/- neurons similarly exhibited waving and swelling of the plasma membrane at axon initial segments and nodes of Ranvier. By contrast, the levels of ankyrin G and voltage-gated sodium channels at these sites, which are significantly reduced in Σ1Σ6-/- mice, were substantially recovered in Σ1-/- mice. We conclude that the truncated βIV-spectrin isoform Σ6 plays a specific role in clustering voltage-gated sodium channels, whereas it is dispensable for membrane stabilization at axon initial segments and nodes of Ranvier.

The spectrin membrane skeleton is a polygonal cytoskeletal meshwork attached to the cytoplasmic face of the plasma membrane (1). The basic unit of the spectrin skeleton is a heterotetramer of two ␣-spectrin and two ␤-spectrin proteins. ␤-Spectrin has an N-terminal actin-binding domain, and the spectrin tetramers are bound to one another indirectly via short actin filaments to form the meshwork. ␤-Spectrin also binds to a membrane adaptor protein ankyrin via the spectrin repeat 15 (2), thereby allowing the attachment of the spectrin-actin meshwork to the plasma membrane. Two major roles are known for the spectrin skeleton. One is to physically stabilize the plasma membrane integrity. In hereditary diseases with mutations in the erythrocyte-specific ␣Iand ␤I-spectrin genes, the erythrocyte membrane becomes fragile, resulting in elliptocytosis and spherocytosis, and eventually hemolytic anemia (3). The second role is to cluster specific integral membrane proteins at high density in specialized regions of the plasma membrane. For example, ␣IIand ␤II-spectrins stabilize the clustering of Na ϩ /K ϩ -ATPase at cell-cell contact sites in polarized epithelial cells through ankyrin B-mediated interaction with Na ϩ /K ϩ -ATPase (4).
Axon initial segments (AIS) 2 and nodes of Ranvier (NR) are specialized axonal domains of the neuron that play essential roles in the firing and amplification of the action potential. The AIS is a short axonal segment located adjacent to the cell body. Voltage-gated sodium channels (VGSC) are highly concentrated at this site and open in response to neurotransmittertriggered excitatory post-synaptic potential, leading to the firing of the action potential (5,6). Axons in vertebrate neurons are often wrapped with myelin sheaths produced by oligodendrocytes in the central nervous system and by Schwann cells in the peripheral nervous system to increase the conduction velocity of the action potential. The NR is a short naked part of the axon between the myelin sheaths (7,8). VGSC are also clustered at NR and open in response to the action potential that has been conducted through the myelinated part of the axonal membrane. In myelinated axons, therefore, the action potential is amplified each time it passes NR so that it does not attenuate before arriving at the axon terminal.
␤IV-Spectrin, one of the five ␤-spectrin family members in mammals, is a component of the spectrin skeleton at AIS and NR (9). At these sites, ␤IV-spectrin binds ankyrin G, an AISand NR-specific isoform of ankyrin (10,11). Ankyrin G in turn binds to VGSC (12). Analyses of mutant mice lacking ␤IV-spectrin have revealed several roles of the spectrin skeleton at AIS and NR. First, the levels of ankyrin G and VGSC are reduced at AIS and NR in ␤IV-spectrin-deficient mice, indicating that through binding to ankyrin G, ␤IV-spectrin stabilizes the clus-tering of VGSC at these sites (11). Second, the plasma membrane at NR is swollen, resulting in abnormal length and width of NR in the absence of ␤IV-spectrin (13). Therefore, the spectrin skeleton physically stabilizes the plasma membrane at NR. Finally, cytoplasmic vesicles are often accumulated around the nodal region in ␤IV-spectrin-deficient mice (13). Although the nature of these vesicles is unknown, this phenotype suggests a role for ␤IV-spectrin also in vesicular transport through the axon. Consequently, loss of the ␤IV-spectrin function causes various neurological defects such as deafness, tremor, and muscle contraction in mice (11,13,14).
The classical ␤-spectrins consist of the N-terminal actinbinding domain, 17 tandem spectrin repeats, a variable region that is not conserved among different ␤-spectrins, and the C-terminal pleckstrin homology domain. For ␤IV-spectrin, cDNAs encoding six splice isoforms (⌺1-⌺6) have been reported (9,11,15). Whereas ⌺2, ⌺3, ⌺4, and ⌺5 have not been shown to be expressed as proteins, an N-terminal-truncated isoform ⌺6, lacking the actin-binding domain and the first 10 spectrin repeats, is expressed at much higher levels than the full-length isoform ⌺1 (11). ⌺6 does not fit in the classical ⌺1-based meshwork of the spectrin skeleton, as it does not bind actin and is only about half the length of ⌺1. This is a specific feature to ␤IV-spectrin among the five mammalian ␤-spectrins, because such a high level expression of truncated isoforms has not been found in other ␤-spectrins (1). Therefore, the presence of the ⌺6 isoform in ␤IV-spectrin suggests a unique membrane skeleton at AIS and NR. Although an essential role for ⌺1 in stabilizing the plasma membrane integrity at NR has been shown using mutant mice lacking ⌺1 but not ⌺6 (16), the role for ⌺6 has not been investigated.
To study the role of the truncated ␤IV-spectrin isoform ⌺6, we generated mice expressing ⌺6 but not ⌺1 (⌺1 Ϫ/Ϫ mice) and performed a side-by-side comparison of their phenotypes with those of previously generated mice lacking both ⌺1 and ⌺6 (⌺1⌺6 Ϫ/Ϫ mice; Ref. 11). If a defect in ⌺1⌺6 Ϫ/Ϫ mice was to be rescued or ameliorated in ⌺1 Ϫ/Ϫ mice, it could be attributed to the loss of ⌺6 function. We demonstrate that ⌺6 is specifically required for the clustering of VGSC at AIS and NR.

EXPERIMENTAL PROCEDURES
Derivation and Genotyping of Mutant Mice-The ␤IV-spectrin genomic locus was isolated from a 129/Sv genomic library using 5Ј regions of cDNAs for ⌺1 and ⌺6. The ⌺1-specific targeting vector was constructed using a neo expression cassette (PGKneolox2DTA; Ref. 17). This targeting vector replaces a 4.6-kb genomic fragment containing exon 1 and part of exon 2, flanked by 1.6-kb PstI-PstI and 4.3-kb ApaI-ApaI genomic sequences derived from the cloned genomic DNA. The construct was electroporated into 129/S4-derived AK7 embryonic stem cells (17), and colonies were selected with G418. Homologous recombination events were screened by the PCR as described (17), using primers corresponding to the neo gene and a genomic sequence outside the targeting construct. Southern blotting was performed according to standard procedures as described (17). Tissue culture and blastocyst injections were performed as described (18). PCR genotyping of mice was performed on tail biopsy samples as described (17), using the fol-lowing combination of primers: forward, 5Ј-AGGCTCTGAT-GTATGGGTGT-3Ј; reverse, 5Ј-ACTCCCACACTGCCACT-CAC-3Ј; and neo, 5Ј-GGATTGGGAAGACAATAGCAG-3Ј.
Electron Microscopy-Mice were perfused with 2.5% glutaraldehyde and 4% paraformaldehyde in 100 mM phosphate buffer, pH 7.4. Optic nerves were removed, cut into small pieces, immersed in the same fixative for 1 h, and postfixed with 1% osmium tetroxide in phosphate buffer for 1 h. Fixed specimens were dehydrated in a graded series of ethanol and embedded in Quetol-812 (Nissin EM, Tokyo, Japan). Ultrathin sections were cut, counter-stained with uranyl acetate and lead citrate, and observed in an electron microscope (H-7500, Hitachi, Tokyo, Japan) at an accelerating voltage of 80 kV.

RESULTS
Organization of the Mouse ␤IV-Spectrin Gene-Schematic structures of the ␤IV-spectrin isoforms, ⌺1 and ⌺6, are depicted in Fig. 1A. Using mouse cDNA fragments corresponding to the N-terminal region of each of the two iso-forms, we isolated genomic DNA fragments encompassing the exons encoding the 5Ј region of the mRNAs for ⌺1 and ⌺6. The organization of these exons was determined by sequence analysis. Exon 1 of the ␤IV-spectrin gene corresponded to the 5Ј non-coding region of ⌺1 mRNA, and the translation initiation codon for ⌺1 was located in exon 2 (Fig.  1B). The ATG codon encoding methionine 1321 of ⌺1 in exon 19 is most likely to be the translation initiation codon for ⌺6 because it is the most upstream ATG in ⌺6 cDNAs (11). We previously identified three ⌺6 cDNAs with different 5Ј non-coding regions (11). Sequences corresponding to these non-coding regions were all found between exons 18 and 19 (Fig. 1B, 18b, 18c, and 18d), suggesting that the three ⌺6 mRNAs are transcribed from different promoters located between exons 18 and 19.
Generation of ⌺1 Ϫ/Ϫ Mice-To disrupt the expression of ⌺1 without affecting that of ⌺6, we constructed a gene targeting vector that replaces a 4.6-kb genomic sequence encompassing exon 1 and part of exon 2 (including the translation initiation codon for ⌺1) with a neo cassette (Fig.  1C). Following electroporation of the construct into 129/S4-derived mouse embryonic stem cells and selection with G418, nine of ϳ100 colonies scored positive for homologous recombination by a PCR screening. Homologous recombination in these clones was confirmed by genomic Southern blotting (Fig. 1D). Germline chimeras were derived from one of the clones and crossed to C57BL/6J mice to derive mutants. Heterozygous offspring did not display an overt phenotype. 3 Heterozygous mutant mice were intercrossed to derive ⌺1 Ϫ/Ϫ mice. Genotyping of the offspring by PCR using tail DNAs as templates (Fig.  1E) showed that the homozygous mice were recovered according to Mendelian expectations. 3 To examine the expression of ⌺1 and ⌺6 in the homozygous mutants, brain lysates from wild-type and ⌺1 Ϫ/Ϫ mice, as well as from a ⌺1⌺6 Ϫ/Ϫ mouse (11), were immunoblotted with anti-␤IV-spectrin antibody. This antibody recognizes the variable region of ␤IV-spectrin, which is present in both ⌺1 and ⌺6 (Fig. 1A). The expression of ⌺1 was completely abolished in the ⌺1 Ϫ/Ϫ brain ( Fig. 2A, top). By contrast, the level of ⌺6 expression in ⌺1 Ϫ/Ϫ brain was comparable with that in the wild type ( Fig. 2A, top), indicating that ⌺6 is not destabilized in the absence of the ⌺1-based classical spectrin meshwork.
Both ⌺1 and ⌺6 Form a Complex with ␣II-Spectrin in the Brain-␣II-Spectrin is most likely the heterotetramerization partner for ␤IV-spectrin at the membrane skeleton of AIS and NR, because ␣II-, but not ␣I-, spectrin is expressed in neurons (19). We examined the level of ␣II-spectrin in ⌺1 Ϫ/Ϫ as well as in ⌺1⌺6 Ϫ/Ϫ brain. Immunoblotting of the brain lysates with anti-␣II-spectrin antibody showed that ␣II-spectrin is expressed at comparable levels in wild-type, ⌺1 Ϫ/Ϫ , and ⌺1⌺6 Ϫ/Ϫ brains ( Fig. 2A, bottom). To examine whether ⌺1 and ⌺6 interact with ␣II-spectrin in neurons, we next performed  Open, shaded, and closed boxes represent the 5Ј non-coding region for ⌺1 mRNA, three distinct 5Ј non-coding regions for ⌺6 mRNAs, and protein-coding regions, respectively. The translation initiation codons (ATG) for ⌺1 and ⌺6 are located in exons 2 and 19, respectively. C, restriction map of the ⌺1-specific gene targeting vector, as well as of the wild-type and mutant alleles. Arrowheads (F, R, and neo) indicate the positions of PCR primers used to genotype mutant mice as shown in E. DTA, diphtheria toxin A gene; neo R , neo-resistance gene; A, ApaI; H, HincII; P, PstI. D, genomic DNAs from wild-type (ϩ/ϩ) and ⌺1 heterozygous mutant (ϩ/Ϫ) embryonic stem cells were digested with HincII and examined by Southern blotting using the 1.6-kb PstI-PstI fragment upstream of exon 1 as a probe. The wildtype and mutant alleles provide 10.5-and 1.8-kb fragments, respectively. E, PCR genotyping using tail DNAs from wild-type (ϩ/ϩ), ⌺1 heterozygous (ϩ/Ϫ), and ⌺1 homozygous (Ϫ/Ϫ) mutant mice as templates. This PCR amplifies 581-and 523-bp diagnostic fragments for wild-type and mutant alleles, respectively. co-immunoprecipitation experiments. ␣II-Spectrin was immunoprecipitated from brain lysates of wild-type, ⌺1 Ϫ/Ϫ , and ⌺1⌺6 Ϫ/Ϫ mice, then the precipitates were immunoblotted with anti-␤IV-spectrin antibody. From wild-type lysate, ␣II-spectrin precipitated both ⌺1 and ⌺6 (Fig. 2B, top). No such co-precipitation was detected with the ⌺1⌺6 Ϫ/Ϫ lysate, indicating that they are not nonspecific signals (Fig. 2B, top). From the ⌺1 Ϫ/Ϫ lysate, ␣II-spectrin co-precipitated an amount of ⌺6 similar to that precipitated from wild-type lysate (Fig. 2B, top). These results suggested that full-length isoform ⌺1 indeed forms the spectrin meshwork with ␣II-spectrin, and more importantly, that the truncated isoform ⌺6 is also able to form a complex with ␣II-spectrin in an ⌺1-independent manner.
Neurological Disorders in ⌺1 Ϫ/Ϫ Mice-Throughout this study, wild-type, ⌺1 Ϫ/Ϫ , and ⌺1⌺6 Ϫ/Ϫ mice were all examined on the same 129/S4 x C57BL/6J F1 hybrid genetic background. Like ⌺1⌺6 Ϫ/Ϫ mice (11), ⌺1 Ϫ/Ϫ mice exhibited tremor that was first observable at weaning. However, it was apparently milder than that of ⌺1⌺6 Ϫ/Ϫ mice. As they become older (usually Ͼ8 months old), ⌺1⌺6 Ϫ/Ϫ mice often exhibit continuous contraction of the hindleg muscle (Ref. 11; supplemental Fig. S1). This phenotype was never observed in more than 50 ⌺1 Ϫ/Ϫ mice, which were older than 8 months (supplemental Fig. S1). In addition, ⌺1⌺6 Ϫ/Ϫ male mice are sterile (11). This is most likely due to a neurological defect that affects sexual behavior, because the histology of their reproductive tissues (testis and epididymis) as well as the morphology and movement of isolated sperms appeared normal, and the mutant mice showed no sign of copulation with wild-type females as judged by the presence of vaginal plugs. 3 When four ⌺1 Ϫ/Ϫ male mice were crossed with wild-type females, by contrast, they all exhibited normal fertility. Overall, the gross phenotypes observed in ⌺1⌺6 Ϫ/Ϫ mice were rescued or milder in ⌺1 Ϫ/Ϫ mice, indicating that ⌺6 is essential for normal neuronal functions.
⌺1-independent Localization of ⌺6 to AIS and NR-Using the antibody against the variable region of ␤IV-spectrin (Fig.  1A), we examined the localization of ⌺6 in ⌺1 Ϫ/Ϫ mice by immunofluorescence staining of tissue sections. ⌺6 localized to AIS in the cerebral cortex and hippocampus (Fig. 3, B and D), as well as in other regions, 3 of ⌺1 Ϫ/Ϫ brain. As expected from the much higher expression level of ⌺6 than ⌺1 in the brain (Ref. 11; Fig. 2A), the level of anti-␤IV-spectrin staining was not significantly reduced in ⌺1 Ϫ/Ϫ mice compared with that in wild-type mice (Fig. 3, A-D). Also in ⌺1 Ϫ/Ϫ sciatic nerve axons, ⌺6 localized normally to NR, although the level of anti-␤IV-spectrin staining was lower than in wild-type neurons (Fig. 3, E and F). These results suggested that the localization of ⌺6 to AIS and NR is regulated by a ⌺1-independent mechanism.
Membrane Destabilization at AIS in ⌺1 Ϫ/Ϫ Mice-Although ⌺6 localized to AIS, the anti-␤IV-spectrin staining pattern of AIS was different between wild-type and ⌺1 Ϫ/Ϫ neurons both in the cerebral cortex and in the hippocampus. Whereas wildtype neurons were smoothly and uniformly stained, ⌺1 Ϫ/Ϫ neurons exhibited waving AIS (Fig. 3, B and D). In addition, staining levels were uneven in different regions of individual AIS (Fig. 3, B and D, arrowheads). The same wavy phenotype was observed when ⌺1 Ϫ/Ϫ AIS in the cerebral cortex were stained with antibodies against ankyrin G and VGSC (Fig. 5, BЈ FIGURE 2. Interaction of ⌺1 and ⌺6 with ␣II-spectrin. A, immunoblot analysis of brain lysates from wild-type, ⌺1 Ϫ/Ϫ , and ⌺1⌺6 Ϫ/Ϫ mice with anti-␤IV-spectrin (top) and anti-␣II-spectrin (bottom) antibodies. The positions of ⌺1 and ⌺6 are indicated. B, ␣II-spectrin was immunoprecipitated from brain lysates of wild-type, ⌺1 Ϫ/Ϫ , and ⌺1⌺6 Ϫ/Ϫ mice, and the immunoprecipitates were immunoblotted with anti-␤IV-spectrin (top) and anti-␣IIspectrin (bottom) antibodies. Bands indicated by asterisks are nonspecific, as they were not consistently detected in the total lysate and anti-␣II-spectrin immunoprecipitate. and EЈ). These results suggested that the AIS membrane, which is defined by the presence of ⌺6, is destabilized in the absence of ⌺1, and therefore that the full-length isoform ⌺1 is essential for maintenance of membrane stability at AIS. Membrane Destabilization and Vesicle Accumulation at NR in ⌺1 Ϫ/Ϫ Mice-The membrane architecture of NR in the central and peripheral nervous systems has been studied using electron microscopy in ␤IV-spectrin-deficient mice (13,16). We compared the ultrastructure of NR in optic nerves of our ⌺1⌺6 Ϫ/Ϫ and ⌺1 Ϫ/Ϫ mice in similar experiments. As reported, the NR plasma membrane of ⌺1⌺6 Ϫ/Ϫ and ⌺1 Ϫ/Ϫ mice exhibited a similar wavy and swollen phenotype (Fig. 4, B and C, compare with the normal membrane in Fig. 4A). Moreover, the fuzzy electron-dense undercoating of the nodal membrane in wild-type mice (Fig. 4AЈ, arrow) was absent in ⌺1⌺6 Ϫ/Ϫ as well as ⌺1 Ϫ/Ϫ mice (Fig. 4, BЈ and CЈ, arrows). These similar defects in ⌺1⌺6 Ϫ/Ϫ and ⌺1 Ϫ/Ϫ NR suggested that also at NR, ⌺1 but not ⌺6 plays an essential role in stabilizing the membrane integrity.
The other nodal phenotype in ␤IV-spectrin-deficient mice is the accumulation of cytoplasmic vesicular structures around the swollen NR membrane (13,16). However, whether the loss of ⌺6 is involved in this phenotype is unclear because the frequency of the appearance of such vesicles has not been compared between mice lacking both ⌺1 and ⌺6 and those lacking only ⌺1. We therefore quantified the proportion of vesicleassociated NR (Fig. 4D, left), as well as the number of vesicles found around a single NR (Fig. 4D, right), in wild-type, ⌺1 Ϫ/Ϫ , and ⌺1⌺6 Ϫ/Ϫ optic nerves. For each genotype, 45 NR from three mice (15 NR from each) were examined. These parameters were not significantly different between wild-type and ⌺1 Ϫ/Ϫ NR, but were much higher in ⌺1⌺6 Ϫ/Ϫ NR. These results suggested that the formation of cytoplasmic nodal vesicles in ⌺1⌺6 Ϫ/Ϫ optic nerves is due largely, if not totally, to the lack of ⌺6.
Also at AIS of Purkinje and granular neurons in the cerebellum, the levels of ankyrin G and Na v 1.6 in ⌺1 Ϫ/Ϫ mice (supplemental Fig. S2, B, BЈ, E, and EЈ) were lower than in wild-type (supplemental Fig. S2, A, AЈ, D, and DЈ) but higher than in ⌺1⌺6 Ϫ/Ϫ mice (supplemental Fig. S2, C, CЈ, F, and FЈ). Na v 1.6 is the major isoform of the VGSC ␤-subunit at AIS and NR (11, 20 -22). Here we detected Na v 1.6 because the anti-pan-sodium channel antibody did not efficiently stain AIS of cerebellar neurons. 3 The elongation of ankyrin G-and VGSC-positive axonal regions, however, was not observed in Purkinje neurons in ⌺1⌺6 Ϫ/Ϫ mice (supplemental Fig. S2).
Localization of VGSC at NR in ⌺1 Ϫ/Ϫ Mice-We next examined the localization of Na v 1.6 at NR in sciatic nerve axons. The anti-pan-sodium channel antibody also failed to stain the NR at detectable levels. 3 The number of Na v 1.6-positive NR was reduced in ⌺1 Ϫ/Ϫ mice (Fig. 6BЈ) compared with wild-type mice (Fig. 6AЈ). However, it was higher than that in ⌺1⌺6 Ϫ/Ϫ mice (Fig. 6CЈ). To quantify the reduction of the Na v 1.6-positive NR in the mutants, the number of Na v 1.6-positive NR in the sciatic nerve sections was counted under the microscope. Counting in nine randomly chosen fields (0.2 mm 2 each) from three mice of each genotype revealed that in ⌺1 Ϫ/Ϫ and ⌺1⌺6 Ϫ/Ϫ axons, the number is reduced to 62 Ϯ 17 and 34 Ϯ 8% (mean Ϯ S.D.), respectively, of that in wild-type axons (Fig. 6D). We were not able to examine the level of ankyrin G at NR because all of the tested anti-ankyrin G antibodies exhibited high nonspecific staining of the myelin sheath. 3

DISCUSSION
Whereas analyses of ␤IV-spectrin ⌺1-specific knock-out mice have revealed an essential role for the full-length isoform in stabilizing the membrane integrity at NR (16), the role and significance of the truncated isoform ⌺6 have been totally unclear. Through a side-by-side comparison of ⌺1⌺6 Ϫ/Ϫ and ⌺1 Ϫ/Ϫ mice on the same genetic background, we here demonstrate that ⌺6 plays a specific role in the clustering of VGSC at AIS and NR.
⌺6 at AIS of ⌺1 Ϫ/Ϫ cerebral and hippocampal neurons exhibited a wavy and uneven staining pattern with anti-␤IVspectrin (Fig. 3). The same pattern was observed when AIS were stained for other AIS-specific membrane proteins, ankyrin G and VGSC (Fig. 5). These results suggest that in the absence of ⌺1, the AIS membrane is ruffled and the cytoplasmic face of the membrane is unequally attached with the putative membrane skeleton composed of ␤IV-spectrin ⌺6 and its binding partner ␣II-spectrin (Fig. 2). Unlike in the cerebral and hippocampal neurons, the abnormal staining pattern of AIS with anti-␤IV-spectrin, antiankyrin G, and anti-VGSC antibodies was not apparent in cerebellar Purkinje neurons in ⌺1 Ϫ/Ϫ mice (supplemental Fig. S2). The reason for this difference is unclear. One possibility might be that sensitivity of membrane stability to the loss of ␤IV-spectrin ⌺1 varies among the AIS of different neurons.
As reported for other ␤IV-spectrin-deficient mice (13,16), the electron-dense fuzzy cytoplasmic undercoating was missing from the NR membrane of optic nerve axons also in our ⌺1⌺6 Ϫ/Ϫ and ⌺1 Ϫ/Ϫ mice (Fig. 4). This was accompanied by swelling of the NR membrane in these axons (Fig. 4). Taken together, the abnormal membrane architecture at AIS and NR of ⌺1⌺6 Ϫ/Ϫ and ⌺1 Ϫ/Ϫ mice suggests that ⌺6 is dispensable for the stabilization of membrane integrity and that the classical spectrin meshwork formed by the full-length ␤IV-spectrin ⌺1 and ␣II-spectrin plays an essential role. This is consistent with the situation in erythrocytes. The erythrocyte-specific ␤I-spectrin, which does not have an abundantly expressed truncated isoform like ␤IV-spectrin ⌺6 (23), is capable of stabilizing the membrane integrity in these cells.
In optic nerves of ⌺1⌺6 Ϫ/Ϫ mice, vesicular structures were often accumulated in the axonal cytoplasm in close proximity to NR (Fig. 4). The frequency of the appearance of such vesicles in ⌺1 Ϫ/Ϫ mice was much lower than that in ⌺1⌺6 Ϫ/Ϫ mice and was similar to that in wild-type mice (Fig. 4). These results indicate that the phenotype is largely rescued by the expression of ⌺6, thus suggesting a ⌺6-specific function. A previous study showed that also in ⌺1 Ϫ/Ϫ mice, vesicles are often accumulated around NR in sciatic nerves (16). However, such accumulation of nodal vesicles was not demonstrated in optic nerves (16). Therefore, the degree of the phenotype might differ between the central and peripheral nervous systems. The identity of the vesicles is currently unknown. One possibility is that they are transport vesicles that deliver specific membrane proteins to NR and that their access to or fusion with the NR membrane is perturbed in the absence of ⌺6. It is important in future studies to elucidate the relevant normal cellular process, impairment of which causes the vesicular phenotype in ␤IV-spectrin-deficient neurons.
The levels of ankyrin G and VGSC were reduced to some extent at AIS and NR in ⌺1 Ϫ/Ϫ mice compared with wild-type mice (Figs. 5, 6, and supplemental Fig. S2). However, they were higher than those in ⌺1⌺6 Ϫ/Ϫ mice (Figs. 5, 6, and supplemental Fig. S2). These results suggest an essential role for ⌺6 in stabilizing the membrane protein cluster containing ankyrin G and VGSC at these sites. In addition, the ankyrin G-and VGSCpositive AIS region in ⌺1⌺6 Ϫ/Ϫ cerebral cortex neurons is approximately twice as long as that in wild-type neurons (Fig.  5). Again, the length of the AIS region was normal in ⌺1 Ϫ/Ϫ neurons (Fig. 5). One explanation for this phenotype is that in the absence of ⌺6, the destabilized membrane protein cluster diffuses toward the axon terminal from AIS. These results therefore further support the role of ⌺6 in maintaining the clustering of VGSC at AIS and NR. A similar phenotype has been reported for Purkinje neurons in cerebellum-specific ankyrin G-deficient mice. In these mice, the AIS region positive for neurofascin 186, an AIS-specific cell adhesion molecule, is elongated at variable levels (24). In our ␤IV-spectrin ⌺1⌺6 Ϫ/Ϫ mice, however, this phenotype was not apparent in Purkinje neurons (supplemental Fig. S2). The reason for this discrepancy is unclear. How intensely this membrane skeletal function depends on ␤IV-spectrin ⌺6 and ankyrin G may vary among neural cell types.
The gross neurological defects observed in ⌺1⌺6 Ϫ/Ϫ mice, such as the tremor, hindleg contraction, and impaired sexual behavior in male, were ameliorated or not observed in ⌺1 Ϫ/Ϫ mice (see "Results" for details). This was correlated with the levels of ankyrin G and VGSC at AIS and NR, which were higher in ⌺1 Ϫ/Ϫ than in ⌺1⌺6 Ϫ/Ϫ mice (Figs. 5, 6, and supplemental Fig. S2). On the other hand, it was not correlated with the level of membrane destabilization at AIS and NR, which was similarly observed in ⌺1 Ϫ/Ϫ and ⌺1⌺6 Ϫ/Ϫ mice (Figs. [3][4][5]. Therefore, the neurological disorders in ⌺1⌺6 Ϫ/Ϫ mice are likely to be due to the reduced level of VGSC at AIS and NR that probably results in insufficient firing and amplification of the action potential, rather than to the membrane fragility at these specific neuronal domains. Co-immunoprecipitation experiments showed that ⌺6 interacts with ␣II-spectrin (Fig. 2). In the spectrin heterotetramer, ␣and ␤-spectrins bind to each other in two ways. The antiparallel lateral association is mediated by the binding of the first two spectrin repeats of ␤-spectrin with the last two spectrin repeats of ␣-spectrin (25,26). The head to head association is mediated by the interaction between the last incomplete spectrin repeat of ␤-spectrin with the N-terminal region of ␣-spectrin (27,28). As ⌺6 harbors the last spectrin repeat but not the first two, it may form a head to head heterodimer with ␣IIspectrin at AIS and NR. However, how ⌺6 fits in the membrane skeleton with ␣II-spectrin at AIS and NR is unclear. It is also unknown why such a truncated isoform, besides the full-length isoform ⌺1, is required for the high density clustering of VGSC. These questions remain to be addressed in future studies. Nonetheless, the results presented here demonstrate an essential role for the unique truncated isoform of ␤IV-spectrin in neuronal function.