High densities of sodium channels at nodes of Ranvier permit action potential conduction and depend on βIV spectrins, a family of scaffolding proteins linked to the cortical actin cytoskeleton. To investigate the molecular organization of nodes, we analyzed qv3J“quivering” mice, whose βIV spectrins have a truncated proline-rich “specific” domain (SD) and lack the pleckstrin homology (PH) domain. Central nodes of qv3J mice, which lack βIV spectrins, are significantly broader and have prominent vesicle-filled nodal membrane protrusions, whereas axon shape and neurofilament density are dramatically altered. PNS qv3J nodes, some with detectable βIV spectrins, are less affected. In contrast, a larger truncation of βIV spectrins in qv4J mice, deleting the SD, PH, and ankyrinG binding domains, causes βIV spectrins to be undetectable and causes dramatic changes, even in peripheral nodes. These results show that quivering mutations disrupt βIV spectrin retention and stability at nodes and that distinct protein domains regulate nodal structural integrity and molecular organization.
Nodes of Ranvier and axon initial segments (AISs) are characterized by high-density clusters of voltage-gated Na+ (Nav) channels that are essential for the generation and propagation of action potentials in myelinated nerve fibers. The formation and stabilization of channel clusters depends on both cellular and molecular mechanisms. For example, deletion of the cytoskeletal scaffolding protein ankyrinG (AnkG) from axon initial segments results in a failure to cluster Nav channels at these sites (Zhou et al., 1998), and mutant animals with disrupted paranodal attachment of myelin have broadened Nav channel clusters at nodes of Ranvier with reduced densities of channels (Dupree et al., 1999; Boyle et al., 2001; Rasband et al., 2003a; Rios et al., 2003). A variety of other proteins, including AnkG-binding cell adhesion molecules [e.g., neuron-glia related cell adhesion molecule (NrCAM) and Neurofascin-186], are also thought to be important in maintaining and clustering nodal ion channels (for review, see Poliak and Peles, 2003; Salzer, 2003).
The spectrins are a family of submembranous scaffolding proteins that, together with ankyrins, cross-link membrane proteins and actin filaments into a stable and flexible network (Bennett and Baines, 2001). For example, the erythroid spectrins are necessary to maintain the highly deformable and elastic properties of circulating red blood cells. Thus, mutations in erythroid spectrins result in red blood cell membrane instability, spherocytosis, and dysfunction (Marchesi and Steers, 1968; Delaunay, 2002). In contrast, the roles of nonerythroid spectrins are not well understood. Recently, a nonerythroid spectrin, βIV spectrin, was found specifically localized to nodes of Ranvier and AISs and was proposed to link Nav channels to the actin-based cytoskeleton through AnkG (Berghs et al., 2000; Komada and Soriano, 2002). Consistent with this idea, Nav channel clusters in βIV spectrin-null mice are disrupted at nodes and AISs (Komada and Soriano, 2002). Thus, βIV spectrin may be important for nodal membrane structure and the stabilization of Nav channels.
To identify the molecular mechanisms underlying these scaffolding functions of βIV spectrin at nodes of Ranvier, we examined here “quivering” (qv) mice with mutations affecting the C-terminal region of βIV spectrin splice variants. A variety of qv alleles exist, each with varying degrees of modification to the predicted primary amino acid sequence of βIV spectrins and each showing a phenotype consistent with both sensory and motor neuropathy (Parkinson et al., 2001). We show here that mice with the most conservative mutation (qv3J) have disrupted nodes of Ranvier in the CNS but mostly normal nodes in the PNS. These results indicate that the C-terminal region is required for the nodal retention and stability of βIV spectrins, which are necessary for maintenance of nodal Nav channel clusters and nodal membrane integrity. In addition, these results suggest that additional βIV spectrin-dependent mechanisms exist in the PNS that can attenuate the phenotype resulting from the mutation found in qv3J mutant mice.
Materials and Methods
βIV spectrin mutant mice. We obtained C57BL/6J-Spnb4qv3J/+ mice from The Jackson Laboratory (Bar Harbor, ME). The mouse line was maintained by heterozygote intercrosses in the animal facility of the University of Connecticut Health Center. The mouse βIV spectrin gene includes 36 exons, and the βIV∑1 splice variant has a predicted length of 2559 amino acids. The qv3J allele contains a single-base insertion at exon 31 and causes a frame shift at amino acid G2209 (Parkinson et al., 2001). A 551 bp fragment at exon 31 including the site of mutation was amplified by PCR (5′ primer, AGGCAGCGCCTTTGCTGCGTC; 3′ primer, TCCTGGTCACAGAGGTCCTTA). StyI (New England Biolabs, Beverly, MA) was used to distinguish the genotype of the pups. Optic nerve and sciatic nerve tissue from control and qv4J mutant mice was kindly provided by Dr. Bruce Tempel (University of Washington, Seattle, WA).
Gait analysis. The gait analysis method was modified from de Medinaceli et al. (1982) and Ozmen et al. (2002). Mice were tested in a confined walkway 10 cm wide by 30 cm long with a dark shelter at the end. Mice were trained several times to walk into the darkened compartment. After dipping their hindpaws into ink on a sheet of Parafilm, mice walked down the corridor on white paper. The distance between each step or between the first and the fifth toes could be compared on the paw prints.
Antibodies. The anti-Nav channel antibodies used included a pan-specific Nav channel antibody that recognizes all neuronal isoforms (Rasband et al., 1999), a mouse monoclonal anti-Nav1.2, and a polyclonal anti-Nav1.6 (Rasband et al., 2003b). The N-terminal-directed anti-βIV spectrin (anti-βIV NT) antibody was generated against a synthetic peptide corresponding to amino acids 15-38 of the βIV∑1 splice variant. The following antibodies have been described previously: rabbit polyclonal and mouse monoclonal anti-Kv1.2 antibodies (Rhodes et al., 1995; Bekele-Arcuri et al., 1996), rabbit polyclonal and mouse monoclonal anti-Caspr antibodies (Rasband and Trimmer, 2001; Rasband et al., 2003b), and a rabbit anti-βIV spectrin “specific” domain (SD) antibody (Berghs et al., 2000). The following antibodies were purchased: anti-ErbB2 antibodies (Santa Cruz Biotechnology, Santa Cruz, CA), anti-CNP (2′,3′-cyclic nucleotide 3′-phosphodiesterase), anti-β actin, anti-α tubulin, and anti-medium neurofilament (NF-M; Sigma, St. Louis, MO), anti-light neurofilament (NF-L) and anti-heavy neurofilament (NF-H; Chemicon, Temecula, CA), and anti-ankyrinG (Zymed, San Francisco, CA). Anti-MBP was a gift from Dr. Elisa Barbarese (University of Connecticut Health Center, Farmington, CT) (Barbarese et al., 1977). Polyclonal antibodies were affinity purified and tested for specificity by Western blot and immunofluorescence with and without blocking by a molar excess of the immunizing peptide.
Immunohistochemistry. The immunostaining method is as described previously (Rasband et al., 1999). Briefly, optic nerves and sciatic nerves from qv mutant and wild-type (WT) mice were dissected immediately after the animals were killed. Nerves were fixed with ice-cold 4% paraformaldehyde in 0.1 m phosphate buffer (PB), pH 7.2, for 30 min. Nerves were then transferred to ice-cold 20% sucrose solution in 0.1 m PB overnight (in the case of qv4J nerves, the tissue was frozen in 20% sucrose and stored at -80°C). For cryosectioning, nerves were frozen in Tissue-Tek (Miles, Elkhart, IL) OCT mounting medium. Sections were cut in 5-μm-thick (optic nerve) and 10-μm-thick (sciatic nerve) sections, placed in 0.1 m PB, spread on gelatin-coated coverslips, and allowed to air dry. The sections were permeabilized for 1 hr in 0.1 m PB containing 0.3% Triton X-100 and 10% goat serum, pH 7.4 (PBTGS). Sections were incubated overnight with primary antibodies diluted to appropriate concentration in PBTGS. Sections were thoroughly rinsed three times in PBTGS (5 min each), followed by application of fluorescently labeled secondary antibody for 1 hr at room temperature. Secondary antibodies were Alexa 488-conjugated goat anti-mouse/anti-rabbit and Alexa 594-conjugated goat anti-rabbit/anti-mouse antibodies (Molecular Probes, Eugene, OR). Finally, labeled sections were rinsed three times in PBTGS, 0.1 m PB, and 0.05 m PB for 5 min each and mounted on slides. Digital images were collected on a Axioskop 2 (Zeiss, Thornwood, NY) fluorescence microscope fitted with a Hamamatsu (Bridgewater, NJ) ORCA-ER camera. In some instances, a Z-stack of images was collected at 0.2 μm intervals, and the resulting stacks were then deconvolved by iterative restoration using the software package Volocity (Improvision, Lexington, MA). The length of each Nav channel cluster was measured using Openlab software (Improvision).
Immunoblotting. Mouse membrane homogenates were prepared from freshly dissected brains. Each brain was homogenized in ice-cold 0.32 m sucrose, 5 mm sodium phosphate, pH 7.4, and 1 mm sodium fluoride, containing 1 mm phenylmethylsulfonyl fluoride, 2 μg/ml aprotinin, 1 μg/ml leupeptin, 2 μg/ml antipain, and 10 μg/ml benzamidine (10 ml/gm wet brain weight). Crude homogenates were then centrifuged at 600 × g for 10 min to remove debris and nuclei. The resulting supernatant was then centrifuged at 45,000 × g for 60 min. This pellet was then resuspended in 2.5 ml of ice-cold homogenization buffer per gram of brain used. Protein concentration was determined using the BCA protein assay kit (Pierce, Rockford, IL). Crude brain homogenates were then diluted in reducing sample buffer to a final concentration of 1 μg/μl, and either 20 or 5 μg membrane proteins were loaded and separated by 6-15% SDS-PAGE. Size-fractionated proteins were then transferred to nitrocellulose membranes and probed according to the procedure described previously. Peroxidase-conjugated goat anti-mouse/anti-rabbit IgGs were used for detection with chemiluminescent reagents (PerkinElmer Life Sciences, Wellesley, MA).
Electron microscopy. Three pairs of 3- to 4-month-old [postnatal day 83 (P83), P110, and P125] qv3J mutants and age-matched littermate WT mice were anesthetized with 0.11 ml of sodium pentobarbital by intraperitoneal injection and then perfused with 2% paraformaldehyde and 2% glutaraldehyde in 0.08 m PB, pH 7.4, containing 0.004% calcium chloride. Optic nerves and sciatic nerves were dissected out and postfixed overnight. These nerves were then osmicated, stained, dehydrated, and embedded in Epon. Ultrathin sections (50-70 nm) of both longitudinal and transverse sections were made for poststaining. Electron micrographs were made using a Jeol (Peabody, MA) JEM-100CX electron microscope. Measurements were performed using NIH ImageJ software (available online at http://rsb.info.nih.gov/nih-image/) and Openlab software. Average ± SD values are given.
Electrophysiology. Compound action potentials (CAPs) from six sciatic and six optic nerves from three pairs of WT and qv3J mice (3-5 months old) were recorded using suction electrodes as described previously (Rasband et al., 1999).
Identification and phenotype of qv3J mutant mice
βIV spectrin is a large, alternatively spliced cytoskeletal protein (six βIV spectrin alternative splice variants have been reported: ∑1-∑4, (Berghs et al., 2000); ∑5, (Tse et al., 2001), and ∑6 (Komada and Soriano, 2002), with diverse protein-protein and protein-lipid interaction domains (Parkinson et al., 2001). To investigate the function of these domains, we chose the qv3J mutant mouse with the most conservative mutation resulting in the smallest change to the predicted primary amino acid sequences of the βIV spectrin splice variants. The qv3J mouse has a single point mutation within the C-terminal SD (Fig. 1A, arrowhead) (Berghs et al., 2000; Parkinson et al., 2001); this domain has also been called the “variable region” by other investigators (Komada and Soriano, 2002), resulting in a novel 49 amino acid extension and lacking the pleckstrin homology (PH) domain (Parkinson et al., 2001). Of the reported βIV spectrin splice variants, this mutation is predicted to affect only βIV∑1, βIV∑3, and βIV∑6 (the domain structures of βIV∑1 and the mutant qv3J βIV∑1 are shown in Fig. 1A).
Because homozygous male qv3J mice are infertile, heterozygous mutants were mated and screened for homozygous qv3J offspring using a PCR-based strategy. Because the StyI restriction enzyme cuts only at the site of the single base pair insertion, it was used to identify mice with the mutant allele(s) (Fig. 1B). The genotypes of homozygous qv3J mice (hereafter denoted qv3J) were verified by immunostaining sciatic nerve using antibodies directed against an epitope located in the SD domain of βIV spectrin (anti-βIV SD) (Berghs et al., 2000) distal to the point mutation; anti-βIV SD and anti-Pan Na+ channel (Pan Nav) immunostaining of sciatic nerve showed that all nodal βIV SD immunoreactivity was lost in qv3J mice (Fig. 1C).
Phenotypically, qv3J mice have progressive neurological and motor impairment and a shortened lifespan; the majority of qv3J mice died before 5 months of age. Heterozygous mice are normal; all subsequent experiments reported here were performed using WT and homozygous qv3J mice. When held by the tail, qv3J mice clasp their hindlegs together rather than in a splayed position like control littermates (Fig. 1D). Gait analysis of young animals (2-3 months old) at the onset of the overt quivering behavior showed early signs of ataxia, including limb weakness and dragging of the hindlegs (Fig. 1E). Older animals (4-5 months) had much more severe ataxia, including paralysis, decreased locomotion, and pronounced quivering (Fig. 1E).
Proteins associated with myelinated axons
Because the qv3J phenotype is consistent with demyelinating neuropathy and/or axonal degeneration, WT littermate and qv3J brain membranes were assayed by immunoblot for changes in the amount of proteins associated with myelinated axons or for the specific nodal, paranodal, and juxtaparanodal domains of myelinated axons (Fig. 1F). For brain Na+ channels, there was no difference in the total pool, but the Nav1.2 subtype was slightly increased in qv3J mice. In contrast, the amount of Caspr (a paranodal protein) (Peles et al., 1997) was reduced in qv3J animals compared with WT, but the levels of Kv1.2 (a juxtaparanodal protein) (Wang et al., 1993) were unchanged. Immunoblotting with antibodies against the myelin proteins MBP and CNP showed no significant differences. The cytoskeletal proteins β-actin and α-tubulin were also unchanged. As expected (Berghs et al., 2000), immunoblots using anti-βIV SD antibodies showed two proteins of ∼250 and ∼140 kDa corresponding to βIV∑1 and βIV∑6 spectrins, respectively, but the lane containing qv3J brain membranes had no immunoreactivity. Although Nav1.2 is typically associated with unmyelinated axons (Gong et al., 1999; Boiko et al., 2001), it has been shown to be increased in animals with inflammatory and genetic hypomyelination or demyelination (Westenbroek et al., 1992; Craner et al., 2003; Rasband et al., 2003b). The reduction in Caspr could be related to a loss of paranodal structures and has been suggested as an early indicator of demyelination (Salzer, 2003).
qv3J mice have aberrant CNS nodes of Ranvier
Myelination and the various axolemmal domains were examined by immunofluorescence (Fig. 2) and electron microscopy (Fig. 3) to determine whether the qv3J mutation results in dysmyelination, altered node of Ranvier formation, and/or axonal degeneration. Immunostaining of 3-month-old qv3J optic nerves with antibodies against Caspr (green) and Nav1.6 (red), the main adult nodal Nav channel (Caldwell et al., 2000; Boiko et al., 2001), showed that many Nav1.6-labeled qv3J nodes were longer than WT nodes (Fig. 2, compare A, arrowhead and B, inset). Nodal Nav1.6 clusters in qv3J mice were on average twice the length of WT clusters and were significantly wider (Table 1). When Nav1.6 channel clusters were increased in length, the immunoreactivity appeared granular and less intense (Fig. 2B, arrow and inset, and D, arrow). In addition, paranodal Caspr staining was often substantially shorter in length (Fig. 2B, inset). Although immunoblot analysis of brain membranes showed an increase in Nav1.2 protein, we did not detect Nav1.2 at optic nerve nodes in qv3J mice (data not shown). In both WT and qv3J mutant mice, Kv1.2 immunoreactivity (green) usually appeared in the juxtaparanode (Fig. 2C,D). In a few instances in which Nav1.6 staining (red) was present in elongated clusters, Kv1.2 immunoreactivity extended to the node without a paranodal gap (data not shown).
In the peripheral nervous system, immunostaining of qv3J sciatic nerve nodes of Ranvier for nodal Nav1.6 (green) and paranodal Caspr (red) (Fig. 2, compare E, F), or Nav channels (green) and juxtaparanodal Kv1.2 (red) (Fig. 2, compare G, H) showed no significant change in the localization of these proteins. Occasionally (much less than 1% of nodes), Kv1 channels were detected in nodal regions (similar to what we observed frequently in other qv mutants; see below). No difference in the length or width of nodal Nav1.6 clusters was detected by immunofluorescence (Table 1).
Electron microscopic analysis of longitudinal optic nerve sections showed a similar increase in node length in qv3J mice (Table 1; Fig. 3, compare A, WT and B, qv3J; arrows delineate the nodal gap). Close examination of paranodal structures revealed that axoglial junctions were normal; all loops directly apposed the axolemma and transverse bands were clearly detected (Fig. 3C, arrows). However, 52% (n = 79) of qv3J nodes had striking membrane protrusions (Fig. 3D, arrows delineate the ends of paranodes). These structures were often filled with vesicles, debris, and mitochondria. Similar protrusions were not seen in the optic nerves of WT mice.
In contrast to the CNS, electron microscopic analysis of PNS nodes of Ranvier confirmed that there was no difference in the length of WT and qv3J nodes (Table 1). Longitudinal sections showed that the majority of qv3J nodes had no significant ultrastructural changes, but 5 of 19 (26%) qv3J peripheral nodes showed some degree of nodal membrane protrusion (Fig. 3, compare E, WT and F, qv3J, arrow). In both WT and mutant qv3J mice, the nodal gap was filled with Schwann cell microvilli. Similarly, cross sections through nodes showed that microvilli were still in contact with the nodal membrane and appeared unchanged. However, the nodal axolemma appeared irregular in qv3J mice, and, in some cases, there appeared to be an accumulation of vesicles (Fig. 3, compare G, WT and H, qv3J, arrow). Together, these results suggest that the SD and PH domains are critical for βIV spectrin function and/or stability and ultimately for the integrity of nodal membrane structure. In addition, the effects of the qv3J mutation appeared to be more dramatic in the CNS than PNS, suggesting that additional mechanisms exist in the PNS that can attenuate the consequences of βIV spectrins lacking the SD and PH domains.
Axon shape and cytoskeletal organization are altered in qv3J mice
Cross sections of optic nerves and sciatic nerves showed that, as for WT mice, myelin was appropriately compacted in the qv3J mutant (Fig. 4, compare A, WT and B, qv3J; sciatic nerve not shown). However, myelinated qv3J optic nerve axons were highly convoluted compared with the normal, more cylindrical shape (Fig. 4B,D, inset); cross sections of qv3J sciatic nerves showed no significant difference in shape (data not shown). To quantify the difference in shape between optic nerve axons in WT and qv3J mice, we calculated a shape factor of S for each axon [S = 4 × π × area/(perimeter)2; S is close to 1 for nearly cylindrical axons, whereas more irregularly shaped axons have values <1]. The qv3J axons were significantly less cylindrical than WT axons (Table 1). The change in shape was even more dramatic for larger fibers (>2 μm2) than for small fibers (Table 1). Finally, we did not observe any axonal degeneration in the optic nerve. These results suggest that βIV spectrin is important not only for nodes but also for membrane shape and structure throughout the axon.
A close examination of the cytoskeleton in longitudinal and cross sections of optic nerve showed that qv3J axons had a large increase in filamentous material compared with WT mice (Fig. 4, compare C, WT and D, qv3J). Higher magnification showed that there was an increase in neurofilament density rather than in microtubule density (Fig. 4, compare E, F, arrows). Immunoblot analysis showed no change in the levels of either actin or tubulin (Figs. 1F, 4G) or in NF-L and NF-H (Fig. 4G). However, NF-M was significantly increased. Thus, the qv3J mutation in βIV spectrin results in widespread cytoskeletal changes in the axons of the optic nerve.
AnkyrinG and mutant βIV spectrin can be detected at nodes in qv3J mice
Double immunostaining using βIV NT and Pan Nav antibodies showed that these proteins colocalized in WT optic nerves (Fig. 5 A, A′), but βIV NT immunoreactivity was absent in qv3J optic nerves (Fig. 5C,C′; C′ shows a single broad Nav channel cluster). Immunoreactivity for AnkG, an adapter protein thought to link Nav1.6 to βIV spectrin, colocalized with Nav1.6 in mutant qv3J optic nerves (Fig. 5, compare B, B′, WT and D, D′, qv3J, arrows). However, the staining intensity of AnkG was reduced and distributed in broader clusters in the qv3J mutant. Thus, although AnkG was detected at CNS nodes, βIV spectrin was not.
In the PNS, βIV spectrin and AnkG immunoreactivities colocalized with Pan Nav staining and were detected at every node of Ranvier in the WT sciatic nerve (Fig. 5E, E′, F, F′). However, the immunostaining for these proteins was significantly attenuated in qv3J mice (Fig. 5G,G′, arrows, and H, H′). Whereas WT nodes had βIV spectrin immunoreactivity that was confined between Caspr-labeled paranodes (Fig. 5 I, I′), 60% (30 of 51) of qv3J nodes had no detectable βIV NT immu-noreactivity (Fig. 5 J, J′), 40% of qv3J nodes had some weak βIV NT staining (Fig. 5 K, K′), and, in one isolated instance, βIV NT immunostaining in qv3J mice was nearly as intense as that seen in WT mice (Fig. 6 L, L′). Importantly, qv3J mouse retinal ganglion cells and cerebellar Purkinje cells did not have βIV spectrin that had accumulated in cell bodies (data not shown). Together, these results suggest that βIV spectrin protein can traffic appropriately to nodes of Ranvier, but, in the absence of the SD and PH domains, it fails to be retained at these sites and may be more rapidly destroyed.
Compound action potentials are mostly normal in qv3J mice
CAPs recorded from sciatic nerves isolated from 3- to 5-month-old WT and qv3J mice showed that the shape of the CAP (Fig. 6A), the calculated peak conduction velocities (Fig. 6B) (n = 6), and the absolute refractory periods (data not shown) were not significantly different at either 25°C or 37°C. These results are not surprising given the lack of changes in ion channel localization and clustering in the qv3J PNS. However, despite dramatic changes at nodes of Ranvier in the CNS, the amplitudes and shape of optic nerve CAPs from qv3J mice were only slightly different from WT mice (Fig. 6A). In some recordings, the second peak of the optic nerve CAP was larger in qv3J optic nerves. This second, slower peak may be attributable to more fibers conducting action potentials at a slower velocity. However, because optic nerve CAPs result from the sum of many thousands of individual action potentials, it is difficult to draw definitive conclusions from changes in amplitudes. Furthermore, the peak optic nerve conduction velocities (calculated from the time to the first peak of the CAP) showed no significant differences (n = 6) (Fig. 6B). Thus, the quivering phenotype may be related to decreased conduction velocities in the CNS. Alternatively, reduced densities of Nav channels at axon initial segments may account for the qv3J phenotype. Consistent with the latter idea, Nav1.6 immunofluorescence staining is reduced at qv3J axon initial segments (Fig. 5M,N, arrowheads).
Loss of the ankyrinG binding domain disrupts PNS nodes of Ranvier
To determine whether the phenotype observed in the qv3J mouse was a specific consequence of the truncation and loss of the SD and PH domains, respectively, we examined nodes of Ranvier in another qv mutant mouse harboring the qv4J allele (Parkinson et al., 2001). These mice have a point mutation that results in a premature stop codon in the 10th spectrin repeat. As a consequence, the reported βIV spectrin splice variants βIV∑1, βIV∑3, and βIV∑6 are predicted to be 60, 8, and 3% of normal length, respectively. The qv4J mutation deletes the AnkG binding domain (Komada and Soriano, 2002). As with qv3J mice, qv4J mice are ataxic, have a reduced lifespan, exhibit deafness (Parkinson et al., 2001), and have disrupted CNS nodes of Ranvier (data not shown). In contrast to qv3J PNS nodes of Ranvier, which appeared normal by light microscopy (Fig. 2F,H), qv4J PNS had a dramatic reorganization in the localization of nodal, paranodal, and juxtaparanodal proteins. Whereas WT mice had nodal Nav channels (red) flanked on each side by paranodal Caspr (green) (Fig. 7A) or juxtaparanodal Kv1.2 K+ channels (green) (Fig. 7C), qv4J mice had many disrupted and apparently degenerating nodes of Ranvier with prominent Nav channel immunoreactivity along the axon and disrupted paranodal junctions (Fig. 7B, arrow). Normal appearing nodes of Ranvier with clustered Nav channels flanked by Caspr were also present (Fig. 7B, arrowhead). Importantly, immunostaining using anti-Caspr (red) and βIV NT antibodies (green) showed that truncated βIV spectrin was undetectable at qv4J nodes (Fig. 7A, inset, WT and B, inset, qv4J), consistent with the previously reported reduction in levels of βIV spectrin mRNA in these mutant mice (Parkinson et al., 2001). Normal appearing nodes often had aberrant Kv1.2 immunostaining that overlapped with Nav channels and extended into outer paranodal zones (Fig. 7D, arrows, overlap is yellow). In contrast to other mutant mice with disrupted paranodes (e.g., Caspr-null and contactin-null) (Bhat et al., 2001; Boyle et al., 2001), Kv1.2 did not extend all the way through the paranode, suggesting that it is only the outermost axoglial junctions that are disrupted and that Kv1.2 targeting to specific domains is altered.
Double immunostaining nodes of Ranvier with Pan Nav (red) and ErbB2 antibodies (green), the latter having been described previously in Schwann cell microvilli (Kim et al., 2002), showed that ErbB2 immunoreactivity was present in a well defined and compact halo around Nav channels in WT mice (Fig. 7E,F). Importantly, qv3J mice also had focal ErbB2 immunoreactivity that surrounded the node of Ranvier (Fig. 7G). In contrast, ErbB2 immunoreactivity in qv4J mice appeared disrupted, was more diffuse, and did not overlap completely with Nav channels (Fig. 7H). These results suggest that the loss of the SD and PH domains may not totally preclude βIV spectrin function and stability, whereas a larger truncation of βIV spectrin, like that found in qv4J mice, is sufficient to disrupt peripheral nodes and Schwann cell microvilli.
The clustering and retention of membrane proteins at nodes of Ranvier is essential for action potential conduction. The mechanisms responsible for this are only now being discovered. For example, several nodal ion channels and cell adhesion molecules, including Nav1.6, Kv3.1b, KCNQ2, NrCAM, and Neurofascin-186, may be clustered at nodes, in part, through their interaction with AnkG (Lambert et al., 1997; Devaux et al., 2003, 2004; Lemaillet et al., 2003). Some studies have suggested that AnkG may be one of the first proteins at nascent nodes, even before Nav channels (Rasband et al., 1999; Jenkins and Bennett, 2002). Furthermore, at the AISs AnkG may coordinate assembly of many of the same proteins present at nodes of Ranvier (Jenkins and Bennett, 2001). These observations suggest that interactions between the cytoskeleton and AnkG may be important for node of Ranvier formation and maintenance. This hypothesis can now be tested because βIV spectrin has been identified as the link between AnkG and the axonal cytoskeleton (Berghs et al., 2000; Komada and Soriano, 2002). Recently, Komada and Soriano (2002) showed that βIV spectrin-deficient mice have reduced densities of Nav channels at both PNS nodes and CNS AISs. In the results described here, we show that, in addition to regulating the levels of nodal Nav channels through AnkG, βIV spectrin also functions to maintain nodal membrane integrity and axon shape in myelinated nerve fibers. Furthermore, by using the qv3J mutant mouse, we show that these properties depend on the C terminus of βIV spectrin, including the SD and PH domains.
The SD and PH domains of βIV spectrin
How does the qv3J mutation disrupt nodes? Truncation of the SD domain and deletion of the PH domain appear not to affect trafficking of the protein but rather its stability and nodal retention. Although the role of the SD domain remains unknown, it is proline rich and may be important for protein-protein interactions. In contrast, PH domains participate in cytoskeleton-plasma membrane adhesion and membrane polarization through their binding to phosphorylated phosphoinositides (Lemmon et al., 2002). One of the most prominent features of the qv3J mutant is the increased node length and the frequent nodal membrane protrusions observed at CNS and PNS nodes. These abnormalities are consistent with a role for βIV spectrin in maintaining nodal membrane structure. βIV spectrin may provide a molecular scaffold not only for proteins but also for the appropriate lipid composition and quantity. Interestingly, a recent study by Nakada et al. (2003) showed that the diffusion of lipids in the AIS membrane was limited. Because the molecular organization of the axon initial segment is similar to the node of Ranvier in many respects (Jenkins and Bennett, 2001), the limited diffusion of phospholipids may be related to their interaction with the PH domain of βIV spectrin. In the absence of βIV spectrin binding to phosphoinositides in the qv3J mutant, the lipid composition may become perturbed, and channels and lipids may be more labile and the overall membrane structure disrupted. Thus, Nav channel retention and membrane stability at nodes of Ranvier may depend not only on protein-protein interactions but also protein-lipid interactions. Based on the observation that Kv1.2 was detected at some disrupted nodes in the qv mutants, we speculate that this regulation may extend even to the kinds of proteins that can partition into the nodal membrane, allowing for the lipid composition to influence the kinds of proteins that can be excluded from or found in the node. Consistent with this idea, Schafer et al. (2004) showed recently that paranodal regions of myelinated fibers have lipid raft-like properties.
Node of Ranvier stability in the PNS
Why is the qv3J mutation more severe in the CNS? The simplest explanation is that nodal integrity is a function of the amount of βIV spectrin that is retained at the node. This conclusion is consistent with the fact that truncated βIV spectrin was undetectable in the CNS of qv3J mice, but low levels of the protein were present at 40% of PNS nodes. Furthermore, βIV NT immunoreactivity was undetectable at PNS nodes of the more severely affected qv4J mice. These observations may reflect slower turnover rates of βIV spectrins in the qv3J PNS compared with the CNS. However, antibody sensitivity combined with different concentrations of spectrins may also underlie the inability to detect diminished levels of βIV spectrin at the CNS node compared with PNS nodes.
Alternatively, a major difference between CNS and PNS nodes is the presence of Schwann cell microvilli. Careful examination of qv3J PNS nodes showed that these structures are normal. A variety of proteins have been described in microvilli (Trapp et al., 1989; Melendez-Vasquez et al., 2001; Scherer et al., 2001; Kim et al., 2002; Goutebroze et al., 2003). Recent experiments in culture have suggested that many of these proteins accumulate at the tips of Schwann cells during early myelination and that microvilli may be involved in Nav channel clustering (Melendez-Vasquez et al., 2001; Gatto et al., 2003). In support of the latter idea, Saito et al. (2003) deleted Schwann cell dystroglycan and found that this results in disrupted microvilli and reduced densities of Nav channels at nodes of Ranvier. Together, these experiments point to a model wherein microvilli are important for the formation of nodes of Ranvier (Salzer, 2003). The analyses of qv3J and qv4J mutant mice are significant because they suggest that microvilli are important not only for node formation but may also contribute to maintenance of mature nodes. With the reduced densities of axonal components like βIV spectrin, it is possible that microvilli can partially compensate by stabilizing the nodal axolemma through as yet unidentified molecular mechanisms. However, when βIV spectrin cannot be detected at nodes, as in the qv4J mice, Schwann cell microvilli are unable to overcome the deficit.
Why are nodes in the PNS of qv4J mutants more disrupted than in qv3J mice? βIV spectrin is alternatively spliced, with six reported variants: ∑1-∑4 (Berghs et al., 2000), ∑5 (Tse et al., 2001), and ∑6 (Komada and Soriano, 2002). Berghs et al. (2000) showed previously that βIV SD immunoreactivity colocalizes precisely with Pan Nav and AnkG immunostaining in peripheral and central nodes of Ranvier, suggesting that βIV∑1, βIV∑3, and/or βIV∑6 may be present at nodes, because each of these splice variants has the SD epitope. However, other data suggested that βIV∑1 and a 140 kDa βIV spectrin (Berghs et al., 2000), likely βIV∑6 (Komada and Soriano, 2002), are the major βIV spectrin splice variants expressed in the nervous system. Our results indicate that at least the βIV∑1 splice variant is present at nodes of Ranvier of qv3J mice, because nodes are also immunolabeled by βIV NT antibodies (Fig. 5A,E). However, we cannot rule out βIV∑6 because these two splice variants cannot be distinguished by immunostaining. The qv4J mutation is predicted to result in βIV∑1 and βIV∑6 proteins that are 60 and 3% of normal length, respectively. It is possible that both splice variants are important for nodes and the nearly complete deletion of βIV∑6 accounts for the more disrupted PNS nodes observed in the qv4J mutant.
A major difference between the qv3J mutation and the qv4J mutation is that the AnkG binding domain is deleted in the qv4J mutant. βIV spectrin may be retained at nodes through the action of both AnkG binding and SD-PH domains. Deletion of both may destabilize spectrins, and thereby the nodes, beyond that which occurs with loss of the PH and SD domains alone. Consistent with this idea, βIV NT immunoreactivity can still be found at some peripheral nodes in qv3J mice but not in the qv4J mutant.
Regulation of neurofilament density by βIV spectrin
An unexpected result from the analysis of the qv3J mutant was a dramatic increase in neurofilament density, an increase in NF-M, and the convoluted shapes of the fibers. Previous studies have shown that axonal diameter is regulated by neurofilament packing (de Waegh et al., 1992; Elder et al., 1999) and that NF-M is the key neurofilament subunit regulating radial axon outgrowth (Garcia et al., 2003; Rao et al., 2003). Although βIV spectrin is thought to be restricted mainly to nodes of Ranvier and axon initial segments in neurons, the results presented here show that the loss of βIV spectrin from nodes can influence the shapes of myelinated axons and the amount of NF-M. Interestingly, in normal PNS axons, there is a reduction in the density of neurofilaments at nodes of Ranvier (Berthold, 1978). Because βIV spectrin is highly enriched at nodes, it is easy to speculate that this protein may act as a negative regulator of NF-M subunits and control neurofilament density. Alternatively, βIV spectrin may be found throughout the axon at levels undetectable by immunofluorescence microscopy, and this non-nodal βIV spectrin may contribute to neurofilament density and overall axon shape. Finally, we cannot rule out the possibility that the qv3J mutation results in a gain of function that increases neurofilament density and expression levels and that the nodal defect is secondary to the disruption of this neurofilament organization.
In summary, the results reported here suggest that both protein-protein and protein-lipid interactions between βIV spectrin, AnkG (and its associated membrane proteins), and the plasma membrane are required for proper cytoskeletal organization and node of Ranvier formation, function, and stability.
This work was supported by the National Institutes of Health (M.N.R. and M.S.), a Wadsworth Foundation Young Investigator Award (M.N.R.), the Alexander von Humboldt Foundation (M.S.), and the American Diabetes Association (M.S.). We thank Dr. Peter Shrager for help and the use of his suction electrode equipment to measure CAPs and peak conduction velocities. We thank Dr. James Trimmer for generously providing antibodies and helpful discussions. We thank Dr. Bruce Tempel for generously providing qv4J tissue. We thank Julie Gross and Maya Yankova for technical advice on electron microscopy.
Correspondence should be addressed to Dr. Matthew Rasband, Department of Neuroscience, University of Connecticut Health Center, 263 Farmington Avenue, Farmington, CT 06030-3401. E-mail:.
Copyright © 2004 Society for Neuroscience 0270-6474/04/249230-11$15.00/0