Anti-GM1 Antibodies Cause Complement-Mediated Disruption of Sodium Channel Clusters in Peripheral Motor Nerve Fibers

Autoimmune lesions at nodal axons in ventral roots

The earliest pathological change in human AMAN is lengthening of nodes in ventral roots (Griffin et al., 1996). Similarly, in L7 ventral roots from two AMAN rabbits at the acute progressive phase, 25 of 52 nodes (48%) were clearly widened under the light microscope compared with the nodes from control rabbit, in particular the paranodal myelin was retracted leaving naked axon. Close examination of lengthened nodes at the electron microscopic level showed disrupted nodal or paranodal structures, including Schwann cell cytoplasm and microvilli (Fig. 1C,D). Paranodal myelin was distorted with small amounts of myelin debris. Some paranodal loops were detached from the axolemma (Fig. 1E).

We previously showed IgG deposition at nodes and axons in the cauda equina (Susuki et al., 2003). To test the idea that the autoantibodies bind to GM1, we first determined the presence of GM1 in these lesions. In ventral roots from the AMAN rabbits at the acute progressive phase, anti-GM1 mAb (GB2) or CT-B stained nodes (Fig. 2A) and axons (supplemental Fig. 1A, available at www.jneurosci.org as supplemental material) with IgG deposition. CT-B staining disappeared during pretreatment of cryosections with chloroform and methanol to extract all glycolipids including gangliosides, suggesting that the stained molecule is GM1 itself and not another glycoprotein bearing the GM1 epitope (data not shown). These findings suggest that the IgG binds to GM1 at nodes and axons. Consistent with this, GM1 epitope was detected at nodal and paranodal axolemma by immunoelectron microscopy (Sheikh et al., 1999). In contrast, there was no or minimal staining of GB2 or CT-B in preserved axons without IgG deposition (supplemental Fig. 1A, asterisks, available at www.jneurosci.org as supplemental material). In control rabbits, we found GB2 stained paranodal myelin, but minimal at nodes (supplemental Fig. 2, available at www.jneurosci.org as supplemental material), similar to previous reports (Molander et al., 1997; Sheikh et al., 1999). In normal nerve fibers, the GM1 epitope may be masked, and autoimmune destruction may be necessary for them to be exposed and allow primary antibody access in immunohistochemistry. Alternatively, GM1 epitope may be lost during the usual immunohistochemical procedures, whereas autoantibodies to GM1 as seen in AMAN may cause them to be stabilized and fixed at the nodes. However, the possibility that GM1 is abnormally localized to the affected nodes or axons cannot be excluded.

We next examined depositions of IgG and C3, the intermediate component of rabbit complement, on ventral roots at the acute progressive phase to show complement-mediated autoimmune attack on nodes and axons. C3 staining was present at the nodes with deposition of IgG (Fig. 2B). Axons were also labeled with both IgG and C3 (supplemental Fig. 1B, available at www.jneurosci.org as supplemental material). These immunohistochemical findings are identical to those in human AMAN, suggesting that autoimmune attack occurs first at nodes and then extends to internodal axons (Hafer-Macko et al., 1996b). C3 deposition was never seen in control rabbits (Fig. 2C), and we did not detect nodal C3 staining in ventral roots from a GalC-sensitized rabbit with demyelinating neuropathy (data not shown). In human AMAN, the disease is usually restricted to peripheral motor nerve fibers, but some patients show sensory nerve involvements as well (Hughes and Cornblath, 2005). Similarly, in a few rabbits sensitized with gangliosides, both motor and sensory nerves are involved (K.S., unpublished observation), but none of the rabbits examined in the present study showed C3 deposition in S1 dorsal roots (data not shown).

Complement can be activated via the classical, alternative, or lectin pathways and culminates in the incorporation of lytic MAC. Some nodes were stained for MAC in human AMAN (Hafer-Macko et al., 1996b). To further implicate the complement-mediated action of anti-GM1 antibodies at the nodes, we examined formation of MAC, the final product of complement activation, in AMAN rabbit nerves. At the acute phase, most nodes with C3 deposition also showed MAC staining (Fig. 2D,E). When we quantified the frequency of the nodal lesions from a total of 1050 nodes observed in ventral roots from three rabbits at the acute progressive phase, 478 nodes (45.5%) had both C3 and MAC staining, 75 (7.1%) had C3 but no MAC staining, and none had MAC but no C3 staining. MAC staining never appeared at nodes in control rabbits (Fig. 2C). To examine the primary lesion of autoimmune attack, we also immunostained the ventral roots with antibodies to CNP, a marker of Schwann cells, or neurofilament-M, a marker of axons. In the apparently lengthened nodes, both C3 and MAC staining was present between the end of two adjacent paranodes (Fig. 2D) and predominantly on the surface of the naked axons (Fig. 2E). These immunohistochemical results together with electron microscopy showing lengthened nodes (Fig. 1C) suggest that the complement-mediated autoimmune attack by anti-GM1 IgG antibodies mainly occurs at nodal axolemma, not myelinating Schwann cells.

We next examined the changes in nodal lesions with complement deposition throughout the clinical course (Fig. 2F). Nodes with C3 deposition were predominantly seen at the acute progressive phase and decreased significantly at the late recovery phase, supporting the pathogenic role of complement on the initial nodal disruption. For the statistical analyses to compare the three groups, acute progressive, early recovery, and late recovery phases, the p value from the global test was <0.001. The results for pairwise comparisons were as follows: acute progressive versus late recovery phases, p < 0.001; acute progressive versus early recovery phases, p = 0.14; early recovery versus late recovery phases, p = 0.017.

Disruption of Na_v channel clusters and related molecules at nodal axons

Next we focused on the change in Na_v channel clusters at the acute progressive phase. Na_v channel localization was altered or disappeared at lengthened nodes with C3 or MAC deposition (Fig. 3A). As the nodal C3 deposition extends, the frequency of disrupted Na_v channel immunoreactivity at the affected nodes significantly increased (Fig. 3B). The frequency of disrupted Na_v channel staining was significantly greater in nodes with MAC deposition than in nodes without MAC staining in AMAN rabbits or in nodes from control rabbits (Fig. 3C). These findings suggest that MAC produced via complement-mediated action of anti-GM1 IgG antibodies first appears at the nodal axolemma, then extends toward the internodes, and disrupts Na_v channel clusters.

Open in a separate window

Figure 3.

Lesions at nodes of Ranvier with complement deposition and disrupted Na_v channel clusters. A, Disruption of Na_v channel clusters at the acute progressive phase. Both C3 (green) and MAC (blue) staining is present at nodes. Clusters of Na_v channels (red) are preserved at lesions restricted to the node (left column), whereas Na_v channels almost disappear at lengthened nodes with extended C3 and MAC staining (right column). Nerve fibers run horizontally in both columns. B, Relationship between Na_v channel cluster disruption and length of complement deposition. Frequency of disrupted Na_v channel immunoreactivity is shown in nodes with C3 staining with lengths of <5, 5–10, or >10 μm. Data were collected from three AMAN rabbits at the acute progressive phase. For the statistical analyses to compare the three groups, the p value from the global test was <0.001. A significant difference was found in each pairwise comparison (p < 0.001). Error bars indicate 95% confidence interval. C, Relationship between Na_v channel cluster disruption and MAC deposition. Frequency of nodes with disrupted Na_v channel clusters is shown in nodes from three control rabbits, in nodes without MAC staining in three AMAN rabbits at the acute progressive phase, and in nodes with MAC deposition. A significant difference was found in the global test (p < 0.001). Specific pairwise comparisons shows that nodes with MAC deposition have more disruption of Na_v channel clusters than nodes without MAC staining or nodes from control rabbits (p < 0.001). There was no significant difference between nodes without MAC staining in AMAN rabbits and nodes from control rabbits (p = 0.95). Error bars indicate 95% confidence interval. DIC, Differential interference contrast image. Scale bars, 10 μm.

To further elucidate how autoimmune attack disrupts Na_v channel clusters, we next examined the molecular organization of nodal axons in ventral roots at the acute progressive phase. The cytoskeletal proteins βIV spectrin and ankyrin G link Na_v channels and NF186 to the actin-based cytoskeleton (Berghs et al., 2000; Komada and Soriano, 2002; Lacas-Gervais et al., 2004; Yang et al., 2004; Sherman et al., 2005). In control rabbits, Na_v channels, βIV spectrin, and NF186 were clearly detected at nodes (Fig. 4A,E). When the complement deposition was restricted to nodes, the clusters of Na_v channels and βIV spectrin were preserved (69.7% of nodes with MAC deposition) (Table 2, Fig. 4B). The disruption of Na_v channel clusters was completely associated with that of βIV spectrin (30.3%) (Table 2). In lesions with C3 deposition extending to paranodes, clusters of Na_v channels and βIV spectrin often showed altered shape and decreased intensity (Fig. 4C). In some lesions, the staining of βIV spectrin did not colocalize with that of Na_v channels (Fig. 4C), indicating the disruption of the Nav channel complex at nodes. This finding was observed in 11 of 208 (5.3%) nodes with MAC deposition in AMAN rabbits but never in control rabbits. The intensity of C3 staining was reduced wherever Na_v channel immunoreactivity was retained (Fig. 4C, arrows). Lesions with more extended C3 were devoid of Na_v channels and βIV spectrin (Fig. 4D). A similar pattern of disrupted NF186 clusters was observed in ventral roots from AMAN rabbits (Fig. 4F–H). These results demonstrate that anti-GM1 antibody-mediated autoimmune attack disrupts the molecular architecture of nodal axons, including loss of Na_v channels.

Open in a separate window

Figure 4.

Alterations of Na_v channels and related molecules at nodal axons by autoimmune attack. All panels show longitudinal sections of ventral roots, and nerve fibers run horizontally. Sections are immunostained with antibodies to C3 (green, all panels), Na_v channels (blue, A–D), βIV spectrin (red, A–D), or pan-NF (red, E–H). Anti-pan-NF antibody specifically stains NF186 in nodal axons. Different stages of nodal disruption in AMAN rabbits at the acute progressive phase are depicted. A, E, Staining results in the control rabbit showing the Na_v channel (A), βIV spectrin (A), and NF186 (E) clusters at the nodes. No C3 staining is detected. B, F, Lesions with C3 staining restricted at nodes. The clusters of Na_v channels (B), βIV spectrin (B), and NF186 (F) are mostly preserved and colocalize with C3 staining. C, G, Lesions with C3 staining extending to paranodes. The node is lengthened (C), and the shapes of clusters of Na_v channels (C), βIV spectrin (C), and NF186 (G) are altered. Note that the staining of Na_v channels does not colocalize with that of βIV spectrin (C). There is abnormal aggregation of Na_v channel staining in the middle of axon (arrowhead). At the region with remaining Na_v channel staining, signal of C3 is relatively weak (C, arrows). D, H, Lesions with remarkably extended C3 staining. The node is lengthened (D), and clusters of Nav channels (D), βIV spectrin (D), and NF186 (H) disappear. DIC, Differential interference contrast image. Scale bars, 10 μm.

Disruption of Schwann cell microvilli

C3 or MAC staining was occasionally detected in the space surrounding nodes. In the peripheral nervous system, Schwann cells form the myelin sheath and extend a collar of microvilli into the nodal gap. Na_v channel clustering requires contact from these Schwann cell microvilli, in which a cell adhesion molecule called gliomedin binds NF186 and initiates node formation (Eshed et al., 2005). These microvilli may play important roles in stabilizing nodal membrane domains (Yang et al., 2004). In human AMAN, Schwann cell microvilli were displaced at the lengthened nodes (Griffin et al., 1996). To investigate whether altered Schwann cell microvilli might contribute to nodal disruption, we analyzed the immunoreactivity of moesin, another protein enriched in microvilli (Scherer et al., 2001) (Fig. 5A), in ventral roots at the acute progressive phase. In lesions with autoimmune attack restricted to the nodal area, MAC staining partially colocalized with moesin immunoreactivity (Fig. 5B). Similar to the axonal molecules at nodes, moesin localization was also disrupted or not present at affected nodes with extended MAC deposition (19.0% of nodes with MAC deposition) (Table 2, Fig. 5C,D). In most nodes with MAC deposition, the disruption of moesin was associated with that of βIV spectrin (Table 2). These findings indicate that the Schwann cell microvilli are also disrupted by autoimmune attack, although it remains unclear whether this occurs primary or secondary to disruption of the nodal axolemma.

Open in a separate window

Figure 5.

Alterations of Schwann cell microvilli by autoimmune attack. All panels show longitudinal sections of ventral roots, and nerve fibers run horizontally. Sections are immunostained with antibodies to MAC (blue), moesin (green), and βIV spectrin (red). Different stages of nodal disruption in AMAN rabbits at the acute progressive phase are depicted. A, Staining results in a control rabbit showing the moesin and βIV spectrin clusters at nodes. No MAC staining is detected. B, Lesions with MAC staining restricted to nodes. The clusters of moesin and βIV spectrin are mostly preserved. MAC staining does not completely colocalize with moesin staining. C, Lesions with MAC staining extending to paranodes. The node is lengthened, and the shapes of clusters of moesin and βIV spectrin are altered. D, Lesions with remarkably extended MAC staining. Node is lengthened, and clusters of moesin and βIV spectrin disappear. DIC, Differential interference contrast image. Scale bars, 10 μm.

Disruption of paranodes and juxtaparanodes

Paranodal junctions may act as a diffusion barrier for ion channel localization (Bhat et al., 2001; Rios et al., 2003). We considered the possibility that disruption of paranodal junctions might lead to alteration of nodal Na_v channels, because Na_v channel disruption was clearly associated with the length of complement deposition along the axons (Fig. 3B). To test this, we immunostained ventral roots at the acute progressive phase with an antibody to Caspr, a cell adhesion molecule at the paranodal axolemma that is essential for the formation of the axoglial junctions (Einheber et al., 1997; Menegoz et al., 1997; Peles et al., 1997; Bhat et al., 2001; Rios et al., 2003) (Fig. 6A). Paralyzed rabbits at the acute progressive phase showed various stages of disruption in nodes and paranodes. In lesions with MAC deposition restricted to nodes, both Na_v channel and Caspr staining was preserved (71.5% of nodes with MAC deposition) (Table 2, Fig. 6B). As MAC deposition extended to paranodes, Caspr staining was reduced or disrupted (Fig. 6C,D). Na_v channel clusters were altered coincident with disruption of Caspr immunoreactivity (22.2%) (Table 2). Despite the altered Caspr staining, 6.3% of Na_v channel clusters appeared normal (Table 2, Fig. 6C). Similar findings were never seen in control animals (Table 2). In lesions with remarkably extended MAC deposition, neither Na_v channel nor Caspr staining was detected (Fig. 6E). These results demonstrate that axoglial junctions are also damaged by the autoimmune attack and suggest that the disruption of paranodal junctions can contribute to alteration of Na_v channel clusters.

Open in a separate window

Figure 6.

Alterations of Na_v channels, paranodes, and juxtaparanodes by autoimmune attack. All panels show longitudinal sections of ventral roots, and nerve fibers run horizontally. Sections are immunostained with antibodies to MAC (blue), Na_v channels (red), and Caspr (green) (A–E), or C3 (green) and Kv channels (red) (F–J). Different stages of disruption in AMAN rabbits at the acute progressive phase are depicted. A, F, Staining in a control rabbit showing the Na_v channel cluster at node, Caspr clusters at paranodes (A), and Kv channel clusters at juxtaparanodes (F). Neither MAC nor C3 staining is detected at node. B, G, Autoimmune lesions restricted to nodes. Clusters of Na_v channels (B), Caspr (B), and Kv channels (G) are not affected. C, H, Autoimmune lesions extended to paranodes. The gap between Caspr staining is lengthened more than that in normal (depicted in A) and is filled with MAC staining (C). Note that the Na_v channel cluster is still preserved, and there are clear gaps between Na_v channel and Caspr clusters (arrowheads). Kv channel clusters are preserved (H). D, I, Autoimmune lesions extended to juxtaparanodes. Node is lengthened, and the clusters of Na_v channels and Caspr are remarkably disrupted (D). Kv channel cluster at juxtaparanode partially colocalizes with C3 staining (yellow) (I). E, J, Lesions with broadly extended MAC or C3 deposition. Node is lengthened, and the staining of both Na_v channels and Caspr completely disappears, whereas these molecules are preserved in adjacent fiber without MAC deposition (E, arrows). Kv channel staining is no longer apparent (J). DIC, Differential interference contrast image. Scale bars, 10 μm.

It is also possible that Kv channels at juxtaparanodes (Poliak and Peles, 2003) (Fig. 6F) redistribute into paranodes and contribute to the acute nerve conduction failure in AMAN. Therefore, we evaluated Kv channel localization in affected nerve fibers. Juxtaparanodal Kv channel staining were preserved until C3 could be detected at these sites (Fig. 6G–I). Kv channel clusters was not apparent in lesions with remarkably extended C3 deposition (Fig. 6J). These results suggest that Kv channel clusters are altered only at the most advanced stage of paranode disruption.

Association of macrophages on nodal disruption

In human AMAN, macrophages recruited by bound autoantibodies and activated complement are thought to overlie and insert processes into the nodal gap, disrupting the myelin sheath attachment sites at paranodes and entering the internodal periaxonal space (Griffin et al., 1996). Macrophages in the periaxonal space with preserved myelin have been confirmed in the spinal nerve roots from AMAN rabbits (Susuki et al., 2003; Yuki et al., 2004). To determine whether macrophages are responsible for initial disruption of nodes, we analyzed macrophage infiltration in ventral roots. Macrophages invaded ventral roots at the acute progressive phase, but most were not localized to the affected nodes with complement deposition (Fig. 7A). Furthermore, nodal molecules were disrupted at lesions with MAC deposition but no overlying macrophages. We quantified the frequency of macrophage binding at the affected nodes with disrupted βIV spectrin or Caspr clusters. Nearly 60% (34 of 59) of nodes did not show associated macrophages. Macrophages were detected at some affected nodes with complement deposition (Fig. 7B), but these cases were rare. The frequency of macrophages at affected nodes with C3 deposition was only 9.0% (47 of 521 at the acute progressive phase). When we examined changes in macrophage invasion over the clinical course, we found that macrophage invasion was significantly more frequent at the early recovery phase (Fig. 7C,D). For the statistical analyses to compare the three groups, acute progressive, early recovery, and late recovery phases, the p value from the global test was <0.001. The results for pairwise comparisons were as follows: acute progressive versus early recovery phases, p < 0.001; early recovery versus late recovery phases, p < 0.001; acute progressive versus late recovery phases, p = 0.42. This suggests a role for macrophages in the clearance of degenerated nerve fibers.

Complement activated by anti-GM1 IgG antibodies disrupts nodes of Ranvier

How do anti-GM1 antibodies induce nodal disruption? It has been suggested that the immune attack directly targets Na_v channel sugar chains containing sialic acid (Takigawa et al., 1995; Waxman, 1995). However, there is no evidence for this idea. IgG from rabbit Bg-16 did not specifically react to Na_v channels highly accumulated at the axon initial segments of cultured rat hippocampal neurons (D. P. Schafer and K.S., unpublished observation). Moreover, the intensity of C3 staining tended to be relatively weak at regions in which some Na_v channel staining was retained, despite remarkable C3 deposition at the adjacent axolemma. Gangliosides are an essential component of lipid rafts and may contribute to maintenance of paranodal junctions and ion channel clusters along peripheral motor nerve axons, regardless of their precise localization at and near nodes. Mutant mice with loss of gangliosides including GM1 by the disruption of β1,4-N-acetylgalactosaminyltransferase gene have altered paranodal junctions, broadened Na_v channel clusters, and aberrant Kv channel localization at the paranodes (Susuki et al., 2007). At the acute phase of the illness in AMAN rabbits, IgG, C3, and MAC depositions were frequently seen at nodes. As in AMAN patients (Hafer-Macko et al., 1996b), complement deposition was most prominent at the acute phase and decreased during recovery. Disruption of Na_v channel staining was associated with MAC deposition. Without localization of macrophages, nodal components may be altered by complement attack, although the effects of macrophages should be examined at more time points. Thus, the complement-mediated immune pathway by anti-GM1 antibodies plays a crucial role for the initial nodal disruption in the AMAN model.

Disruption of axon–Schwann cell interaction in AMAN model

How are Na_v channel clusters disrupted? To answer this question, a better understanding of the mechanisms that stabilize Na_v channels at the nodal axolemma is required. Paranodal junctions may function as a diffusion barrier to restrict the lateral mobility of ion channels (Bhat et al., 2001; Rios et al., 2003) (supplemental Fig. 3A, available at www.jneurosci.org as supplemental material). Consistent with this idea, at abnormally lengthened nodes with MAC formation extending to paranodes, Na_v channel and Caspr clusters were disrupted. The disruption of paranodal junctions may result in myelin detachment and subsequently remove the diffusion barrier leading to altered Na_v channel clusters. However, loss of paranodes is not sufficient to disrupt Na_v channel clusters. Even in mutant mice without paranodal junctions, Na_v channels still cluster at nodes (Bhat et al., 2001; Rios et al., 2003). Another possible explanation is that both the specialized axonal cytoskeleton and Schwann cell microvilli were disrupted in AMAN rabbits. These structures are required for stabilization of Na_v channel clusters (Berghs et al., 2000; Komada and Soriano, 2002; Lacas-Gervais et al., 2004; Yang et al., 2004; Eshed et al., 2005; Sherman et al., 2005) (supplemental Fig. 3A, available at www.jneurosci.org as supplemental material). Thus, the disruption of Na_v channel clusters could be a consequence of destruction of all these structures (supplemental Fig. 3B, available at www.jneurosci.org as supplemental material).

Disruption of Na_v channel clusters was reported in experimental allergic neuritis, an animal model for AIDP (Novakovic et al., 1998). At the onset of disease in the AIDP model, Na_v channels became more diffusely distributed into paranodal and juxtaparanodal zones. In contrast, the diffuse Na_v channel staining along the axon was never seen in the AMAN model. In AMAN, Na_v channel disruption appears to be more crucially involved in pathophysiology, because primary autoimmune attack occurs directly at the nodal axolemma in which Na_v channels aggregate at high density (Hafer-Macko et al., 1996b). This is in contrast to AIDP in which the autoimmune target is myelin (Stoll et al., 1991; Hafer-Macko et al., 1996a). Despite different primary lesions, similar alterations of Na_v channel clusters were observed in both animal models for AMAN and AIDP. That is, Na_v channels become undetectable during the disease, followed by binary Na_v channel immunoreactivity and subsequent fusion of these clusters to form new nodes during recovery. Disrupted molecular architecture of the nodal axons is well known in general different peripheral nerve demyelination models. In the lysolecithin-induced demyelination model, nodal widening or segmental demyelination is associated with altered Na_v channels and paranodes at the peak, and two Na_v channel clusters at heminodes ultimately appear to fuse during remyelination (Dugand$\breve{z}$ija-Novaković et al., 1995; Arroyo et al., 2004; Schafer et al., 2006). Mice with peripheral dysmyelinating or demyelinating phenotypes (e.g., Trembler-J and MPZ-null) display altered paranodes, abnormal Kv channel distribution, or nodes containing Na_v1.8 (Ulzheimer et al., 2004; Devaux and Scherer, 2005). Furthermore, retention of Na_v channel clustering in axons requires the interaction with Schwann cells during development of peripheral myelinated axons (Vabnick et al., 1996; Eshed et al., 2005; Schafer et al., 2006): Na_v channels initially cluster at heminodes that are associated with myelinating Schwann cells and appear to move toward each other until they fuse and form the condensed cluster. Together, these findings suggest that altered interaction between axons and Schwann cells may be the main cause of pathophysiology in AMAN.