Ion channel sequestration in central nervous system axons

Abstract

Reliable and efficient functioning of myelinated axons in the CNS depends upon a highly heterogeneous distribution of ion channels. While considerable information has been published regarding the processes responsible for ion channel sequestration in the peripheral nervous system (PNS), corresponding details for the CNS have only recently become available. Much of the development of these fibres occurs over the first few weeks after birth. In the neonatal rat optic nerve, for example, the density of Na⁺ channels in premyelinated axons is less than 2 μm⁻², and the conduction velocity is only about 0.2 m s⁻¹ (Waxman et al. 1989). Conduction velocity increases slowly over the first 2 weeks, and then rises sharply to the adult value of about 11 m s⁻¹ for the fastest fibres. This rapid rise occurs between postnatal day (P)15 and P25 and corresponds to the onset of myelination and the clustering of voltage-dependent Na⁺ channels at developing nodes of Ranvier (Rasband et al. 1999). Shaker-type K⁺ channels are organized at high density within the juxtaparanodes, axonal regions just beyond the paranodal axoglial junctions, but this sequestration develops almost 1 week later than that of Na⁺ channels (Wang et al. 1993; Rasband et al. 2000). This review will focus on the mechanisms responsible for the redistribution of these two classes of ion channels during development and on the consequences of disruption due to disease.

Clustering of Na⁺ channels during development

Over the past several years, investigators have been intrigued by questions concerning the mechanism through which the location of nodes of Ranvier is specified. Are these sites intrinsically determined by the axon, or do the myelinating glia control nodal spacing? In the PNS there is now strong evidence that Na⁺ channel clustering is initiated by Schwann cells just after the latter become committed to myelination (Vabnick & Shrager, 1998). These channels appear to be excluded from regions of close contact between Schwann cells and axons, accumulating just outside the tips of glial processes (Novakovic et al. 1996). The clusters move laterally as these processes grow, ultimately fusing with a neighbouring cluster to form a node. This system was first described in remyelinating axons (Dugandzija-Novakovic et al. 1995; Tzoumaka et al. 1995), and was later confirmed and extended in developing fibres (Vabnick et al. 1996; Custer et al. 1999) and in vitro (Ching et al. 1999). In the CNS, an alternative scenario has been suggested from experiments on retinal ganglion cells in culture (Kaplan et al. 1997). The number of Na⁺ channel clusters in ganglion cell axons increased significantly when the neurons were suspended above a non-contacting layer of oligodendrocytes. Astrocytes were inactive, but glia-conditioned medium was effective, leading to the hypothesis that a soluble factor released by oligodendrocytes induced clustering at sites predetermined by the axon. Experiments on these same axons in vivo suggest, however, that a mechanism similar to that proposed for the PNS is also at work here.

Rasband et al. (1999) examined Na⁺ channel clustering during development of the optic nerve using immunocytochemistry and electrophysiology. The extent of myelination was gauged by double labelling for Na⁺ channels and either myelin-associated glycoprotein (MAG) or Caspr/paranodin, a component of the axoglial junctions at paranodes (Einheber et al. 1997; Menegoz et al. 1997; Peles et al. 1997). MAG-positive oligodendroglia were first detected at P7 (in contrast to the PNS, MAG is expressed prior to ensheathment; Bartsch et al. 1989). Caspr immunoreactivity was found at the edges of some of these early MAG-labelled processes, but Na⁺ channel clusters were not seen until P9-P10 (Fig. 1a, P9). Clusters became more frequent and more focal over the next 2 weeks (Fig. 1a, P15, P19 and Adult). Similarly, the shape of the compound action potential changed and the conduction velocity also increased over this same period (Fig. 1C). When two zones of Caspr immunoreactivity were seen in close proximity, suggestive of early paranode formation, focal Na⁺ channel immunofluorescence was invariably found in the gap between them (Fig. 2a). Among the sites with only one Caspr-positive zone, about one-third were associated with Na⁺ channel immunoreactivity, which was usually graded in intensity with the strongest signal adjacent to the Caspr region. Importantly, there was never any overlap between these two proteins. A quantitative analysis over the period P5-P60 revealed that the appearance of Caspr-positive sites preceded the rise in frequency of Na⁺ channel clusters by about 2 days (Fig. 1b). A small number of binary Na⁺ channel sites were seen transiently (10 % at P12, decreasing to 1 % at P19), reminiscent of the fusing of clusters in the PNS (Fig. 1a, P15, arrowhead).

Open in a separate window

Figure 1

Clustering of Na⁺ channels during development

All data are from rat optic nerve. A, P9, a broad cluster of Na⁺ channels; P15, P19 and Adult, fields of view (FOV) showing an increase in the number of nodal clusters and overall decrease in cluster length. Note the binary aggregate at P15 (arrowhead). Scale bars, 10 μm. B, the time course (in days) of clustering for Na⁺ and K⁺ channels as well as detection of Caspr-labelled axoglial junctions: Na⁺ channel (NaCh) clusters per FOV, ♦; percentage of NaCh clusters with Kv1.2, □; Caspr sites per FOV, ▪. (Modified from Rasband et al. (1999) with permission from the Journal of Neuroscience and from Rasband et al. (2000) with permission from the Journal of Neurocytology.) C, compound action potentials recorded from P2, P16 and adult rat optic nerves. (Modified from Rasband et al. (1999) with permission from the Journal of Neuroscience.)

Open in a separate window

Figure 2

Ion channel and Caspr localization in wild-type and mutant Shiverer optic nerves

A, Na⁺ channels (green) and Caspr (red) are found exclusively in nodal and paranodal zones, respectively, in wild-type (WT) optic nerve. B, Na⁺ channels (green) and K⁺ channels (red) in WT optic nerve. C, double labelling of Shiverer (Shi) optic nerves for Na⁺ channels (green) and Caspr (red) shows aberrant Na⁺ channel clustering and axoglial junctions (arrowhead and asterisk, respectively). D, diffuse K⁺ channel (red) and infrequent, irregular Na⁺ channel aggregates (green) in Shi optic nerve. Scale bars, 10 μm.

The above combination of association, timing and morphology all point to a mechanism in which Na⁺ channels cluster only after oligodendrocytes adhere and initiate early steps in myelination, including early paranode formation. However, 12 % of the Na⁺ channel clusters seen during development had only weak or undetectable neighbouring Caspr immunofluorescence, leaving some uncertainty. The hypomyelinating mutant mouse Shiverer (Shi) provided a means to test further the mechanisms involved. Shi animals lack myelin basic protein (MBP). As a result, CNS myelination is reduced: many axons are ensheathed, but myelin is uncompacted and paranodes are highly irregular (Rosenbluth, 1980b, 1981; Inoue et al. 1981). There is also a substantial difference in Na⁺ channel expression. In the brain, Na⁺ channel types I and III are normally found primarily in neuronal cell bodies, while type II channels are primarily axonal (Westenbroek et al. 1989; Gong et al. 1999). Localization of Na⁺ channel types I and III is unchanged in Shi mice, but the density of the axonal type II Na⁺ channels is elevated in this mutant (Noebels et al. 1991; Westenbroek et al. 1992). Na⁺ channel types I, II and III, as well as Scn8a/PN4/NaCh6 are all expressed by retinal ganglion cells (Fjell et al. 1997). By labelling MAG, Rasband et al. (1999) demonstrated that despite the hypomyelination, large numbers of oligodendrocytes were present in the developing Shi optic nerve. On the other hand, the number of focal Na⁺ channel clusters was dramatically reduced at all stages, relative to littermate controls. In the adult, Caspr staining was highly abnormal, often appearing in single isolated zones (Fig. 2C). Importantly, even when Caspr-positive regions were of an unusual shape, Na⁺ channel clusters were often associated with them. In some cases, Caspr partially surrounded a Na⁺ channel site, with no overlap in immunoreactivity. This was highly reminiscent of the lakes of particles and irregular axoglial junctions seen in electron micrographs (Rosenbluth, 1981). The Shi results argue strongly for the contact-dependent hypothesis. The 6-fold reduction in frequency of node-like Na⁺ channel clusters in Shi mice corresponded to the irregularity of paranode formation, not to the number of oligodendroglia, even though many of these cells differentiated to the stage of multiple ensheathment. Finally, the reduction in clustering occurred despite the fact that expression of axonal Na⁺ channels was increased overall. In contrast, a recent report suggests that in spinal cord axons of a galactolipid-deficient mouse, Na⁺ channels cluster despite a highly disrupted distribution of Caspr (Dupree et al. 1999). It will be important to analyse this preparation further by double labelling for Na⁺ channels and Caspr, and by looking earlier than the one time point (P30) examined, since the crucial information concerns the period during which the channels cluster. Once sequestered, Na⁺ channel clusters can be stable for at least 1 week even in the face of complete demyelination (Dugandzija-Novakovic et al. 1995).

Clustering of Shaker-type K⁺ channels

In mammals, voltage-dependent K⁺ channels are not present at nodes of Ranvier, but instead are found localized at juxtaparanodal zones with little or no encroachment into paranodal or nodal sites (Fig. 2b) (Chiu & Ritchie, 1981; Wang et al. 1993; Rhodes et al. 1997; Rasband et al. 1998). In the CNS tissues of cerebellum, brainstem and optic nerve, juxtaparanodal K⁺ channels have been shown to be heteromultimeric complexes consisting of Kv1.1, Kv1.2 and Kvβ2 subunits (Wang et al. 1993; Rhodes et al. 1997; Rasband et al. 2000). The function of these channels during normal conduction has been difficult to ascertain. Application of 4-aminopyridine (4-AP) has almost no effect on compound action potentials in the PNS whereas in the optic nerve the action potential amplitude and duration are increased (Gordon et al. 1988; Rasband et al. 2000). On the other hand, in intracellular recordings, the depolarizing afterpotential of rat phrenic axons increases significantly on exposure to this drug (David et al. 1995). Further, it has been suggested that juxtaparanodal K⁺ channels may serve to inhibit repetitive activation of nodal Na⁺ channels (Chiu & Ritchie, 1981). Experiments in the PNS have shown that K⁺ channels may be transiently localized to nodal and/or paranodal zones where they stabilize action potential conduction during development (Vabnick et al. 1999) and inhibit conduction during remyelination (Rasband et al. 1998). Genetic deletion of Kv1.1 causes conduction abnormalities in both the CNS and PNS (Smart et al. 1998; Zhou et al. 1998b).

K⁺ channel localization has been studied following demyelination and during remyelination of PNS nerve fibres to determine the mechanisms of clustering and channel organization (Rasband et al. 1998). These experiments showed that initially, after loss of the overlying myelin sheath, channels were no longer localized to strictly defined zones adjacent to the node of Ranvier, but instead were free to diffuse laterally through the axonal membrane, suggesting a requirement for myelin in stabilizing and retaining juxtaparanodal K⁺ channels. During subsequent remyelination, K⁺ channels were initially detected in the gap between adjacent Schwann cells, but then were redistributed through paranodes into juxtaparanodes. Inhibition of Schwann cell proliferation and remyelination blocked clustering. Vabnick et al. (1999) showed that during developmental myelination in the PNS, K⁺ channel localization followed a sequence of events similar to that seen during remyelination. These observations suggested that, as for Na⁺ channels, myelin is critical for the discrete localization of K⁺ channels. Since the process of Na⁺ channel clustering and the repertoire of juxtaparanodal K⁺ channels are the same between the CNS and PNS (Wang et al. 1993; Rhodes et al. 1997; Rasband et al. 1998, 2000), one might expect that K⁺ channel clustering is also similar in these two tissues. What is the experimental evidence for, or against, this conclusion?

Recently, the mechanisms of Shaker-type (Kv1) K⁺ channel clustering at juxtaparanodes in the CNS have been investigated in the spinal cord and optic nerve during normal developmental myelination, and in dysmyelinating mutant mice (Wang et al. 1995; Dupree et al. 1998; Baba et al. 1999; Rasband et al. 2000). During myelination in the rat optic nerve, the detection of K⁺ channel immunoreactivity and clustering lagged behind that for Na⁺ channels by several days (Fig. 1b). K⁺ channels were first seen at about 2 weeks after birth, and importantly, in contrast to myelination and remyelination in the PNS (Rasband et al. 1998; Vabnick et al. 1999), they were not present at nodes or paranodes, but rather at juxtaparanodes (Fig. 3a). Further, these early K⁺ channel clusters were often found in discrete bands at the paranode-juxtaparanode interface (Fig. 3b), and they alternated with Caspr immunoreactivity (Rasband et al. 2000). These bands have been interpreted as local accumulations of K⁺ channels between the innermost axoglial junctions (Vabnick et al. 1999), and probably correspond to the paranodal lakes of particles described in freeze-fracture studies of myelinated nerve fibres (Fields et al. 1986). These observations indicate that K⁺ channels are excluded from axoglial junctions and suggest that the overlying oligodendrocyte may directly influence channel localization. The number of nodal sites with adjacent juxtaparanodal K⁺ channel staining continued to increase during development until all Na⁺ channel clusters had corresponding K⁺ channel immunoreactivity (Fig. 1b). In the adult rat optic nerve, Kv1.1/Kv1.2/Kvβ2 K⁺ channels were confined exclusively to juxtaparanodal zones, which extended for 3–20 mm. Indeed, there is virtually no overlap among Na⁺ channels, K⁺ channels and Caspr in nodal regions (Fig. 4a).

Open in a separate window

Figure 3

Clustering of K⁺ channels during development

A, Kv1.2 (red) immunoreactivity at P14 is found in juxtaparanodal zones just beyond the paranodal Caspr staining (green). (Modified from Rasband et al. (2000) with permission from the Journal of Neurocytology.) B, at early stages of clustering and development, Kv1.2 (red) is often found in ‘bands’ at the juxtaparanode-paranode interface; Na⁺ channels are present only in nodal clusters (green). Scale bars, 10 μm.

Open in a separate window

Figure 4

Ion channel clustering in the CNS

A, rat optic nerve triple labelled for Kv1.2 (blue), Caspr (red) and Na⁺ channels (green). Scale bar, 10 μm. B, sketches of Na⁺ (NaCh) and K⁺ (K⁺ Ch) channel clustering during development.

In a similar study on K⁺ channel clustering, Baba et al. (1999) examined the mouse optic nerve with special emphasis on the lamina cribrosa, where the transition from unmyelinated to myelinated nerve allows for a specific analysis of K⁺ channel localization and its dependence on myelin within the same axon. They found that K⁺ channels were clustered only in regions that were myelinated; proximal to the lamina cribrosa, channels were diffuse. Further, they reported a delay in channel clustering, similar to that in the rat optic nerve (Rasband et al. 2000), that correlated with late stages of myelination. Together, the developmental studies of K⁺ channel clustering in optic nerve suggest that myelin and glial cells are important for the localization of K⁺ channels in the CNS.

Studies of mice with defects in CNS myelination or myelin structures have been especially informative with respect to the requirement for glial involvement of K⁺ channel clustering. Wang et al. (1995) examined K⁺ channels in the spinal cord and ventral fibre tracts of the brainstem in Shi mutant mice and found both an upregulation in K⁺ channel expression and a redistribution of the channels such that they were no longer found in focal juxtaparanodal aggregates, but were diffusely localized throughout the nerve fibre. However, the authors noted that during early development, juxtaparanodal clusters of K⁺ channels were present in the mutant spinal cord. They concluded from these results that factors other than myelin might be responsible for the initial targeting and clustering of channels, but that in adults, myelin was necessary for channel maintenance in juxtaparanodal zones. This conclusion differs from that of Rasband et al. (1999), who examined the initial localization of K⁺ channels in Shi optic nerve. In contrast to littermate control animals, which had normal nodal Na⁺ channel and juxtaparanodal K⁺ channel clusters (Fig. 2b), they found that at no point were channels present in juxtaparanodes of Shi mice at high density. Instead, in the absence of well-formed axoglial junctions and compact myelin, K⁺ channels were diffusely distributed throughout the axolemma (Fig. 2D). Since the degree of myelination was not assessed in the Shi spinal cord (Wang et al. 1995), the discrepancies in clustering may reflect subtle differences in sensitivity to the defect in the MBP gene between myelination in the optic nerve and spinal cord. This conclusion is further supported by the fact that PNS myelin is only slightly affected by the mutation (Rosenbluth, 1980a).

Examination of optic nerves from the mutant mouse Jimpy (Baba et al. 1999), which has a defect in the proteolipid protein (PLP) gene resulting in severe CNS hypomyelination, showed that there were occasional isolated instances of juxtaparanodal K⁺ channels. Importantly, these cases of clustered channels always overlapped with staining for MBP, suggesting that when myelin was present, channel sequestration was possible. Finally, the galactolipid-deficient mouse has disrupted paranodes, with a greater frequency of heminode formation, increased node length, and a total loss of the transverse bands normally associated with paranodal axoglial junctions (Dupree et al. 1998). When K⁺ channel localization in spinal cords from these mutant mice was examined, channels were no longer confined to juxtaparanodal zones, but were usually found to extend into the paranode and internode (Dupree et al. 1999). Since the primary defect in these animals is altered paranodal axoglial junctions, these results suggest that these specialized sites are important for the maintenance and localization of channels at juxtaparanodes.

The experimental data in large part thus suggest a model (Fig. 4b) wherein oligodendrocytes initially form paranodal structures that are Caspr positive (Fig. 4b (1)). Subsequent to the formation of these sites, Na⁺ channels accumulate at the edges of elongating myelinating processes (Fig. 4B (2)). By immunofluorescence these Na⁺ channel clusters may be detected at the edges of Caspr-labelled axoglial junctions as either very broad (Fig. 4b (3a)) or binary (Fig. 4b (3b)) aggregates. As myelination progresses, the nodal Na⁺ channels condense into more focal (∼1 μm) clusters, and K⁺ channels are first detected in juxtaparanodal zones, or occasionally between axoglial junctions near the paranode-juxtaparanode interface (Fig. 4B(4)). Finally, as paranodal axoglial junctions continue to mature, the detection of K⁺ channels in bands between Caspr-labelled axoglial junctions becomes very rare, with these channels confined almost exclusively to juxtaparanodal zones. Caspr and Na⁺ channels are found only in paranodal and nodal zones, respectively (Fig. 4B (5); see also Fig. 4a).

Molecular participants in the clustering process

While the precise molecular interactions responsible for Na⁺ channel sequestration at nodes of Ranvier are unknown, many recent experiments point to candidate participants. The ankyrins are a set of intracellular proteins thought to serve as a link between integral membrane proteins and the underlying spectrin-based cytoskeleton (Bennett, 1992). AnkyrinB has been shown to associate with brain Na⁺ channels by immunoprecipitation and binding affinity measurements (Srinivasan et al. 1988). A recently discovered gene, known as ankyrinG (Kordeli et al. 1995) or Ank3 (Peters et al. 1995), has multiple isoforms and is expressed in both the nervous system and the kidney. Of particular interest here, Ank3/ankyrinG is found at nodes of Ranvier and axon initial segments, sites of high density of Na⁺ channels. Further, Bennett and his colleagues have shown that neurofascin, L1, NrCAM and NgCAM, a group of cell recognition/adhesion molecules with extracellular Ig-like and fibronectin type III domains, bind ankyrin at their intracellular regions, and are localized to nodes of Ranvier (Davis & Bennett, 1994; Davis et al. 1996). At embryonic stages, these proteins are initially diffusely distributed along axons, and they cluster during the first postnatal week (Lambert et al. 1997). Ank3/ankyrinG is invariably seen adjacent to MAG-positive Schwann cell processes. However, while many clusters of neurofascin and NrCAM are likewise paired with glial processes, in some cases these proteins appear not to be linked to MAG expression. Bennet and co-workers consider the Ank3/ankyrinG binding proteins to be pioneer molecules in node formation. In the developing optic nerve, Rasband et al. (1999) found that Ank3/ankyrinG clustering preceded that of Na⁺ channels, supportive of this view. On the other hand, ankyrin immunoreactivity appeared to be resolvable in two regions, with strong staining in the nodal gap and weaker label extending into paranodal regions. Exploiting a selective expression pattern of Ank3/ankyrinG exons in brain, Zhou et al. (1998a) genetically deleted this protein from the cerebellum. Na⁺ channels, as well as ankyrin, were then absent from Purkinje cell initial segments. The distribution of neurofascin was disrupted, suggesting that at these sites Ank3/ankyrinG may be required for localization, a mechanism different from that proposed for nodes. Finally, from direct measurements of mobility within the surface membrane, Winckler et al. (1999) have shown that transmembrane proteins in initial segments are immobilized largely through cytoskeletal linkages, presumably via ankyrins.

Several other proteins have been considered as strong contenders for a role in ion channel clustering. Tenascin-R (J1-160/180, janusin, restrictin) is present in the extracellular matrix and is concentrated at nodes in the CNS (ffrench-Constant et al. 1986; Bartsch et al. 1993). Further, contactin/F3/F11, a glycosyl phosphatidylinositol-anchored (GPI) protein with homology to the β2 subunit of the Na⁺ channel, is a receptor for tenascin-R, and there is evidence that tenascin-R can associate directly with Na⁺ channels (Pesheva et al. 1993; Isom et al. 1995; Xiao et al. 1996; Srinivasan et al. 1998). Application of a recombinant N-terminal domain of tenascin-R to Xenopus oocytes expressing Na⁺ channels resulted in a significant increase in peak Na⁺ current with no change in kinetics (Xiao et al. 1999). Weber et al. (1999) found that in mice deficient in tenascin-R, the timing of Na⁺ channel clustering was identical to that of wild-type controls, and the density of Na⁺ channels at nodes, as judged by immunofluorescence intensity, was also unchanged. However, the conduction velocity in the optic nerve (but not the sciatic nerve, which lacks tenascin-R) was almost 2-fold lower in the knock-out mouse. A reduction in Na⁺ currents due to the removal of the modulation by tenascin-R that was seen by Xiao et al. (1999), may be responsible for this functional deficit. Thus, tenascin-R may alter Na⁺ channel function at nodes of Ranvier, but appears not to be crucial for localization.

Other possible participants include receptor protein tyrosine phosphatase β (RPTPβ), expressed by glia and some neurons. RPTPβ binds to contactin/F3/F11 and NrCAM through different domains (Grumet, 1997). Since contactin is GPI anchored, its cellular action may be transmitted through laterally associated NrCAM or Caspr/paranodin (Peles et al. 1998). From its timing and pattern of expression, MAG was proposed to be essential for Na⁺ channel clustering in the PNS (Vabnick et al. 1996). However, channel sequestration is normal in MAG-deficient animals (Vabnick et al. 1997). Other proteins with changes in expression coincident with clustering include NCAM, L1, P0, and the p75 nerve growth factor receptor. These have yet to be tested for participation in the clustering process. The expression pattern of different Na⁺ channel subtypes is temporally and spatially regulated in the CNS (Felts et al. 1997b; Gong et al. 1999). There is as yet, however, only minimal information available on the identity of the channel types present at nodes of Ranvier or initial segments. Scn8a/PN4/NaCh6 is the only Na⁺ channel clearly identified at nodes, but this has to date been only in the PNS, and not all nodes are populated by this channel (Novakovic et al. 1999). The recent availability of good subtype-specific antibodies should accelerate work in this area.

In contrast to nodal Na⁺ channels, candidate proteins that may be involved in the clustering of juxtaparanodal K⁺ channels are just beginning to be identified. One potential mediator, Caspr2, a member of the neurexin superfamily, has been shown to colocalize with K⁺ channels at juxtaparanodes in both the CNS and PNS (Poliak et al. 1999). Reciprocal co-immunoprecipitation experiments showed that Kv1.2, Kvβ2 and Caspr2 coassociate. However, the interaction was shown to depend on C-terminal PDZ binding motifs (named after the proteins PSD-95, Dlg and ZO-1) in both Caspr2 and Kv1.2, suggesting that the interaction is indirect and requires a scaffolding protein. Interestingly, an antibody generated against the first two PDZ domains of PSD-95, a protein that has been shown to cause clustering of NMDA receptors and Kv1 K⁺ channels in vitro (Kim et al. 1995; Kim & Sheng, 1996), stains juxtaparanodes in the optic nerve and spinal cord (Baba et al. 1999). This observation suggests that PSD-95, or another member of this large family of PDZ domain-containing proteins (Sheng & Wyszynski, 1997), mediates the interaction between the K⁺ channels and Caspr2. However, positive identification of the interacting protein has not yet been accomplished. It is likely that other components will be found that interact with K⁺ channels at juxtaparanodes. Given their highly specific localization (Fig. 4a), and the dependence of clustering and maintenance of channel aggregates on myelin, one can expect a larger community of proteins that may include both intrinsic neuronal participants that link the channel complexes to the cytoskeleton, and extrinsic glial/myelin proteins that are also part of the juxtaparanodal K⁺ channel complex. This latter interaction may be direct or through other associated proteins such as Caspr2.

A final important question regarding CNS axonal Na⁺ channels concerns their role in demyelinating diseases, e.g. multiple sclerosis. It has previously been demonstrated that in the PNS conduction is invariably blocked when myelin is stripped from an entire internode, but conduction can be restored by just minimal glial ensheathment (Shrager, 1988; Shrager & Rubinstein, 1990). It is likely that this return of function is dependent on the early Na⁺ channel clustering that accompanies initial steps in remyelination (Smith et al. 1982; Dugandzija-Novakovic et al. 1995). Felts et al. (1997a) have now been able to make intracellular recordings from single, identified axons in the rat spinal cord. They showed that conduction can be restored over demyelinated segments ∼2.5 mm in length with just a low level of glial repair. It is not yet known whether Na⁺ channel clustering plays a role in this process, but the results in the PNS, coupled with similarities in node formation in the two systems, would predict that this is the case. Na⁺ channel expression, as measured by saxitoxin binding, is increased in demyelinated lesions in multiple sclerosis (Moll et al. 1991). The K⁺ channel-blocking drug 4-AP has been used to reverse neurological deficits in multiple sclerosis (Bever et al. 1994) since these channels are exposed in demyelinated axons and may even be nodal in remyelinating fibres, while conduction in most normal axons is unaffected by this compound (Kocsis & Waxman, 1980; Sherratt et al. 1980; Bostock et al. 1981; Rasband et al. 1998). Understanding the cellular and molecular details of ion channel sequestration in myelinated axons is thus important in both normal development and disease. This issue is part of the larger area of neuron-glial interactions that govern multiple aspects of nervous system function.

Acknowledgments

References