The motor axonal variant of Guillain–Barré syndrome is associated with anti-GD1a immunoglobulin antibodies, which are believed to be the pathogenic factor. In previous studies we have demonstrated the motor terminal to be a vulnerable site. Here we show both in vivo and ex vivo, that nodes of Ranvier in intramuscular motor nerve bundles are also targeted by anti-GD1a antibody in a gradient-dependent manner, with greatest vulnerability at distal nodes. Complement deposition is associated with prominent nodal injury as monitored with electrophysiological recordings and fluorescence microscopy. Complete loss of nodal protein staining, including voltage-gated sodium channels and ankyrin G, occurs and is completely protected by both complement and calpain inhibition, although the latter provides no protection against electrophysiological dysfunction. In ex vivo motor and sensory nerve trunk preparations, antibody deposits are only observed in experimentally desheathed nerves, which are thereby rendered susceptible to complement-dependent morphological disruption, nodal protein loss and reduced electrical activity of the axon. These studies provide a detailed mechanism by which loss of axonal conduction can occur in a distal dominant pattern as observed in a proportion of patients with motor axonal Guillain–Barré syndrome, and also provide an explanation for the occurrence of rapid recovery from complete paralysis and electrophysiological in-excitability. The study also identifies therapeutic approaches in which nodal architecture can be preserved.
The motor axonal variant of Guillain–Barré syndrome, acute motor axonal neuropathy (Feasby et al., 1986; McKhann et al., 1993; Hughes and Cornblath, 2005) characteristically follows Campylobacter jejuni infection and is associated with serum anti-GM1, -GD1a and -GalNAc-GD1a ganglioside antibodies (Lugaresi et al., 1997; Ho et al., 1999; Ogawara et al., 2000). Gangliosides have diverse functions related to neural development, maintenance and regeneration, including stabilizing the axoglial junction at the node of Ranvier (Sheikh et al., 1999b; Susuki et al., 2007a; Silajdzic et al., 2009). GD1a has been identified in the motor nerve terminal and nodal axolemma (Sheikh et al., 1999a; De Angelis et al., 2001; Gong et al., 2002; Goodfellow et al., 2005), sites that correspond to those predicted from clinical, electrophysiological and pathological data to be affected in motor axonal forms of Guillain–Barré syndrome (Griffin et al., 1996; Ho et al., 1997; Kuwabara et al., 2004).
The distal motor nerve, including nerve terminals and the spinal roots, are relatively accessible to circulating factors that may account for vulnerability of these regions to injury compared with nerve trunks, owing to blood nerve barrier permeability variations (Burkel, 1967; Saito and Zacks, 1969; Olsson, 1990). As such, the rapid clinical recovery seen in some patients with acute motor axonal neuropathy could be due to very distal axonal conduction injury and block, a site with the capacity to readily regenerate (Ho et al.,1997; Goodfellow et al., 2005). Conversely, severe proximal axonal injury resulting in widespread axonal degeneration would inevitably lead to permanent motor axonal deficits, as is seen in some acute motor axonal neuropathy cases (Hiraga et al., 2005a, b).
Anti-ganglioside antibody-mediated mouse and rabbit models of acute motor axonal neuropathy have been generated that focus on sciatic nerve and ventral root axons, or on axonal components of neuromuscular junctions (Susuki et al., 2003; Sheikh et al., 2004; Goodfellow et al., 2005). In a rabbit ventral root model, destabilization of nodal and paranodal structures including loss of voltage-gated sodium channels (Nav1.6) was interpreted as the consequence of antibody and complement-mediated axoglial disruption (Susuki et al., 2007b), which could be protected with a complement inhibitor (Phongsisay et al., 2008).
Complement mediated effects at the nodes of Ranvier in the ventral root mirror those demonstrated in patient autopsy tissue (Hafer-Macko et al., 1996). As the node of Ranvier is vital for impulse propagation, understanding acute motor axonal neuropathy immunopathology at this site is important. The node of Ranvier is organized into three subdomains—the nodal gap, the paranode and the juxtaparanode (Scherer, 1996; Poliak and Peles, 2003). Nav1.6 is expressed at the node (Caldwell et al., 2000) along with the cytoskeletal protein ankyrin G and the cell adhesion molecules neurofascin 186 and neuronal cell adhesion molecule (NrCAM). At the paranode, the axoglial junction is formed by the axolemmal proteins contactin and Caspr, while neurofascin 155 is the glial receptor to this complex. At the juxtaparanode, voltage-gated potassium channels localized on the axon in complex with Caspr 2 and Tag1, play a role in repolarization of the resting membrane potential following an action potential (Wang, 1993). Glycosyltransferase knockout mouse studies indicate that GD1a or related gangliosides modulate the structural and functional integrity of this site, although the precise mechanisms are poorly understood (Sheikh et al., 1999b; Susuki et al., 2007a; Silajdzic et al., 2009).
In our ex vivo mouse model of acute motor axonal neuropathy, motor nerve terminals develop severe functional and pathological injury when exposed to anti-GD1a antibody with complement activation (Goodfellow et al., 2005). In this model GD1a is enriched in axons through using the GD3 synthase knockout mouse and thus becomes more vulnerable to anti-GD1a antibody mediated injury than normal mice. How this relates more precisely to motor axonal GD1a level in man is unknown. This model also requires a heterologous complement source (human serum) in order to achieve the pathological effects, as previously reported (Willison et al., 2008). The pore forming action of complement is central to the development of this injury and that mediated by other anti-ganglioside antibodies, in part through allowing uncontrolled calcium influx into the nerve terminals, with subsequent Ca2+-dependent protease, calpain, activation and cleavage of structural proteins in the axon terminal (O'Hanlon et al., 2003).
This study set out to assess whether anti-GD1a-antibody mediated injury could be observed to occur at the nodes of Ranvier in the distal portions of the axon, upstream from the motor nerve terminal. If present, we also intended to determine the mechanism of action and functional effects of any observed injury that might lead to therapeutic intervention, analogous to our previous approach at the neuromuscular junction.
Materials and methods
Male GD3 synthase knockout mice (GD3s−/−) (Okada et al., 2002) were crossed with B6/Cg-TgN(Thy1-CFP) × Dilute Brown non-Agouti (DBA/1) mice that endogenously express cyan fluorescent protein (CFP) in their axons (Feng et al., 2000), herein termed GD3s−/−/CFP. GD3s−/− mice relatively overexpress GD1a compared with wild-type counterparts, bind anti-GD1a antibody strongly and deposit complement resulting in injury, as reported (Goodfellow et al., 2005). To confirm this in the crossed strain used in this study, we assessed triangularis sterni muscles from both GD3s−/−/CFP and wild-type/CFP mice for anti-GD1a antibody immunoreactivity. GD1a levels were significantly greater in GD3s−/−/CFP compared with wild-type/CFP mice (P < 0.05). Consequently, wild-type mice were found to be relatively insensitive to anti-GD1a antibody mediated injury compared with GD3s−/− mice and the latter were used for all subsequent experiments. Mice aged 6–12 weeks were killed by CO2 inhalation. Experiments complied with UK Home Office guidelines.
Antibodies and reagents
Anti-GD1a IgG2b monoclonal antibody (previously termed MOG35, herein termed ‘anti-GD1a antibody’) was generated as described (Bowes et al., 2002; Boffey et al., 2005). By way of control for the specificity of the anti-GD1a antibody, GalNAc transferase−/− mice that lack GD1a showed no binding or pathological effects from antibody exposure, as previously reported (Goodfellow et al., 2005). Antibodies to channels, other proteins and membrane attack complex, C5b-9 are detailed in the online Supplementary Table 1. Eculizumab, a humanized anti-human C5 monoclonal antibody that binds plasma C5 to prevent membrane attack complex formation, and ALXN3300 (the isotype-matched control monoclonal antibody) were supplied by Alexion Pharmaceuticals (Cheshire, USA). The synthetic peptide AK295 binds calpain I, II and cathepsin B to prevent their activation and proteolytic action (Li et al., 1996). Toxins were used as follows: α-bungarotoxin (Molecular Probes, UK); α-latrotoxin (Alomone Labs, Israel) at 12 nM; tetrodotoxin (Biotium Inc., USA) at 5 μM; and vecuronium (Organon Laboratories Ltd, Cambridge, UK) at 5 µM. Secondary antibodies conjugated to Alexa Fluor 488 and 647 were applied at 1 : 200 dilution.
Ex vivo and in vivo muscle and nerve permeability studies
Triangularis sterni muscle, phrenic and sural nerve were maintained alive in Ringer’s medium (116 mM NaCl, 4.5 mM KCl, 1 mM MgCl2, 2 mM CaCl2, 1 mM NaH2PO4, 23 mM NaHCO3, 11 mM glucose, pH 7.4), pre-gassed with 95% O2/5% CO2 at room temperature (∼20°C). Muscle and nerve (desheathed by slitting and opening the epineurium with a fine needle, or left intact) were incubated with 100 μg/ml anti-GD1a antibody for 2 h at 32°C, 30 min at 4°C and a final 10 min at room temperature, plus α-bungarotoxin to label the neuromuscular junctions. Antibody control preparations were incubated with Ringer’s alone. Preparations were rinsed in Ringer’s prior to fixation in 4% paraformaldehyde (20 min, room temperature). Tissue was then rinsed in phosphate buffered saline, 0.1 M glycine and phosphate buffered saline. Tissue was incubated with anti-IgG2b-488 (1 : 200) and the pan anti-neurofascin antibody NFC2 (1 : 1000) with 0.5% Triton X-100 in blocking solution (1% goat serum and 1% l-lysine) overnight at 4°C. Intramuscular nerve bundles were categorized as follows: single fibres, small bundles (<15 μm), medium bundles (15–35 μm) and large bundles (>35 μm). Nodes of Ranvier were identified by neurofascin staining and the anti-GD1a antibody immunofluorescence quantitated and compared to control tissue. For in vivo studies, triangularis sterni muscle was removed from mice injected i.p. 16 h previously with 3 mg anti-GD1a antibody. Phosphate buffered saline was used for control groups.
For phrenic and sural nerves, isolated nerves were incubated ex vivo with anti-GD1a antibody under intact and desheathed conditions. It was thereby established that desheathing was essential for achieving anti-GD1a antibody binding at nodes of Ranvier in nerve trunks and that under these conditions antibody binding levels were equivalent to intramuscular nerve nodes of Ranvier (data not shown). All studies on nerve trunks were thus conducted on desheathed nerves.
Ex vivo preparations for complement activation and nodal protein disruption
Muscle and nerve preparations were processed as above, with the additional step of incubation with 40% normal human serum (source of complement) for 3 h at room temperature prior to fixation. Ten micrometre cryosections were stained for membrane attack complex, nodal channels and other proteins overnight at 4°C. In order to identify nodes of Ranvier, fluoromyelin green (1: 400), which labels lipids, or dystrophin (1: 200), which labels the myelin sheath, were applied. Secondary antibodies were applied for 3 h at room temperature as follows: anti-rabbit IgG-647 (1 : 300) for Nav1.6, Caspr, NFC2, Kv1.1, neurofilament; anti-mouse IgG1-647 (1 : 300) for ankyrin G, moesin, NrCAM and dystrophin; anti-mouse IgG2a-555 (1 : 200) for membrane attack complex. Nodes of Ranvier with a normal immunostaining pattern for nodal proteins were scored as present or absent/abnormal. To control for the possible confounding effect of synaptic injury that occurs with anti-GD1a antibody application to nerve muscle preparations (Plomp and Willison, 2009), Nav1.6 immunostaining at nodes of Ranvier was compared between antibody treated and α-latrotoxin (2 nM) treated tissue.
To assess the contribution of membrane attack complex to injury, eculizumab (100 μg/ml) was added to normal human serum 10 min prior to incubation with the muscle (Halstead et al., 2008a). To investigate the contribution of calpain, AK295 (100 μM), was added concurrently with normal human serum. AK295 was first assessed in dose ranging studies (25–200 μM) and the lowest fully protective concentration was used. In eculizumab-treated and -unprotected intramuscular axons, the presence of axonal CFP was used to monitor axonal integrity. After AK295 treatment, the intensity of neurofilament immunoreactivity was quantified at the nerve terminal as delineated by α-bungarotoxin staining and compared to AK295-unprotected tissue levels. The efficacies of eculizumab and AK295, as monitored by immunostaining profiles, were expressed as the percentage of protected versus unprotected signals at the relevant node of Ranvier sites.
Perineural and extracellular recordings
Triangularis sterni nerve-muscle preparations were set up for electrophysiological recordings in Ringer’s at room temperature (20–22°C) using 2 M NaCl-filled micro-electrodes with a resistance of 25–45 MΩ. Recordings were made from nerve terminals and small and large intramuscular nerve bundles after anti-GD1a antibody incubation followed by normal human serum for 3 h. Perineural waveforms associated with nerve terminal action potentials were made as previously described (Braga et al., 1991). Muscles were paralysed with 5 µM vecuronium to prevent twitching. In some experiments the same microelectrode was used to measure muscle resting membrane potentials. Signals were amplified, recorded and analysed as per the nerve extracellular recordings below.
For extracellular recordings, nerves were mounted in a Perspex recording block across three chambers and sealed in with vacuum grease. Nerves were stimulated at 1 Hz and supramaximal voltage (Grass S88 stimulator). Signals were amplified (CED1902), digitized (NIDAQ-MX A/D converter, National Instruments, Austin, TX, USA) and analysed using WinWCP version 4.1.0. Phrenic nerves and sural nerves remained in the recording chamber throughout the experiments and recordings were made for 2 h on the application of normal human serum. At termination, 5 µM tetrodotoxin was applied to confirm that the recorded waveform originated from the opening of sodium channels. A representative graph of the positive peak value of compound action potential (CAP) over time was plotted to convey conduction. A minimum of 200 control waveforms were averaged prior to the addition of normal human serum. Absolute CAP values vary between experiments and thus the percentage of the starting CAP peak value was calculated for each of 3–5 preparations and averaged for each treatment group. A 2-sample t-test was used for comparison of phrenic and sural nerve conduction.
Image acquisition and analysis
Fluorescent images were captured on both a Zeiss Axio Imager Z1 with ApoTome attachment and a Zeiss Pascal confocal microscope and analysed using ImageJ software. For quantitation of antibody and membrane attack complex deposition, the fluorescence signal at the region of the node of Ranvier was measured with background fluorescence subtracted. Where relevant, measurements were categorized by bundle size as described above. Quantitation of the neurofilament signal over the motor endplate was performed as reported (Halstead et al., 2008a). Measurements were pooled from three experiments. Non-parametric data are presented as box and whisker plots. Mann–Whitney mean rank test was used to compare possible statistical differences between groups (1% level of significance). For comparison of nodal protein immunostaining, nodes of Ranvier positive for individual markers were counted for each bundle category and the chi-squared test used at a 1% level of significance.
Anti-GD1a antibodies are preferentially deposited at distal motor nerve nodes of Ranvier
Anti-GD1a antibody was applied to ex vivo triangularis sterni muscle preparations from GD3s−/−/CFP mice and its deposition immunolocalized and quantified at the nodes of Ranvier. Anti-GD1a antibody also binds intensely at the motor nerve terminal of GD3s−/− mice (Fig. 1A; upper area of image) as shown in previous studies (Goodfellow et al., 2005). Anti-GD1a antibody deposits are prominent at the nodes of Ranvier of distal intramuscular axons, identified by the nodal and paranodal marker, pan-neurofascin antibody (Fig. 1A). Juxta-terminal nodes of Ranvier bear the most prominent antibody deposits, the fluorescence intensity of anti-GD1a antibody deposits at nodes of Ranvier decreasing with increasing distance from the nerve terminal. In order to quantitatively assess this, intramuscular nerve bundles were categorized into three groups (Fig. 1B). The single arrow indicates a single fibre, double arrow a small bundle (<15 µm) and triple arrow a medium bundle (15–35 µm). A further category of large nerve bundles (not evident in Fig. 1B) was assigned for bundle diameters exceeding 35 µm. Anti-GD1a antibody applied to triangularis sterni muscle preparations ex vivo were deposited at significantly higher levels at single fibre nodes of Ranvier compared to all other bundle categories and control tissue (Fig. 1C and D; P < 0.001). This was also evident for small (P < 0.001) and medium (P = 0.0023) bundles. Large bundles had an insignificant anti-GD1a antibody deposition level, comparable to control tissue, suggesting anti-GD1a antibody was unable to gain access to bundles of this size following topical application. In order to assess anti-GD1a antibody penetration to these intramuscular nerve compartments when delivered through the vascular bed (as opposed to organ bath incubation), anti-GD1a antibody was injected intraperitoneally and 16 h later the triangularis sterni muscle was removed for antibody quantification, as for ex vivo preparations above. Equivalent results to the ex vivo findings were observed, with antibody deposits being greatest in the distal part of the nerve in a gradient-dependent manner when categorized by bundle size (Fig. 1E). In order to establish that these differences were not due to a proximal to distal gradient of GD1a expression at nodes of Ranvier in nerve, frozen sections of permeabilized intramuscular nerve bundles in which antibody access is expected to be uniform were stained with anti-GD1a antibody, and the signal intensity was found to be the same, irrespective of the nerve bundle size (data not shown). These ex vivo and in vivo findings demonstrate that anti-GD1a antibody is able to bind to intramuscular nerve nodes of Ranvier in a distal to proximal downward gradient, presumed due to the relatively increasing impermeability of the blood nerve barrier to antibody as bundle size increases.
Nodal proteins are disrupted and distal motor nerves are rendered inexcitable by anti-GD1a antibody directed complement activation
Membrane attack complex deposition was demonstrated at nodes of Ranvier of the distal intramuscular nerves in response to the addition of an exogenous source of human complement in the form of normal human serum (Fig. 2A). As with anti-GD1a antibody, membrane attack complex deposition as assessed by fluorescence intensity for anti-membrane attack complex antibody was gradient-dependent, with significantly higher levels at single fibre nodes of Ranvier compared to all other categories (Fig. 2B; P < 0.001) and significantly higher levels at small bundle nodes of Ranvier compared to larger categories (Fig. 2B; P < 0.001). Axonal injury was further characterized by the complete loss of the endogenous axonal CFP signal, both at the nerve terminal and along the distal axon as illustrated in Fig. 2C. Even in large bundles, the CFP signal was relatively attenuated in antibody plus normal human serum treated preparations.
In order to assess the functional effect of membrane attack complex deposition to the distal axonal region, electrophysiological assessment of local ion currents was performed by recording perineural currents at the nerve terminal, small and large nerve bundles. In control tissue (anti-GD1a antibody without normal human serum), biphasic waveforms were observed that correspond to currents flowing through Na+ and K+ channels, respectively (Fig. 2D; top panels). After treatment of tissue with anti-GD1a antibody plus normal human serum as a complement source, there was a complete loss of both K+ (broken arrow) and Na+ (solid arrow) current flow at the nerve terminal and nerve bundles, with the exception of preserved Na+ current in large bundles (Fig. 2D; lower panel, arrow).
To investigate for structural alterations, the node of Ranvier was analysed for appearance under phase microscopy and by immunostaining for node of Ranvier proteins located at various nodal sub-domains. The percentage of nodes of Ranvier with intact staining for Nav1.6, the sodium channel isoform expressed at the peripheral nerve nodes of Ranvier (Caldwell et al., 2000) was reduced (<90%) in single fibres and small bundles after treatment, compared to control (Fig. 3A; P < 0.001). The cytoskeletal protein ankyrin G was similarly affected (Fig. 3B; P < 0.001), as was the paranodal axolemmal protein Caspr (Fig. 3C; P < 0.001). Staining for neurofascin was partially lost at single fibre and small bundle nodes of Ranvier (Fig. 3D; P < 0.001 and P = 0.001, respectively), and at nodes of Ranvier where it was preserved, the pattern was disrupted. Immunostaining to the potassium channel Kv1.1 localized to the juxtaparanode was unaffected in all bundle categories (Fig. 3E).
The complete and rapid disappearance of key nodal axolemmal proteins as assessed by immunostaining following 3 h of normal human serum exposure was striking. In order to assess this in more detail for Nav1.6, intermediate stages of dissolution of Nav1.6 immunostaining were qualitatively examined at 15 and 30 min after the addition of complement treatment. At 15 min there was no alteration to staining; however by 30 min a proportion of the nodes of Ranvier developed punctuate and dispersed Nav1.6 staining, indicating fragmentation and spread of Nav1.6 channel clusters bound by the anti-Nav1.6 antibody (Fig. 7D).
In this model, the severe and concomitant motor nerve terminal injury might have more proximal motor axonal consequences. By way of control, Nav1.6 staining at nodes of Ranvier was assessed after α-latrotoxin-induced injury and found to be unaffected (data not shown).
Complement inhibition completely protects nodes of Ranvier from anti-GD1a antibody-mediated injury
In order to demonstrate the role for the membrane attack complex component of complement activation, the C5 complement inhibitor eculizumab, that completely prevents membrane attack complex assembly, was studied. Eculizumab protected Nav1.6, ankyrin G and Caspr immunostaining at nodes of Ranvier from injury mediated by complement activation, compared with the isotype control antibody ALXN3300. In quantitative analysis, the percentage of nodes of Ranvier with intact Nav1.6 staining is significantly greater on the addition of eculizumab compared to the isotype-matched control monoclonal antibody ALXN3300 at single fibres and small bundles (Fig. 4A; P < 0.001). As demonstrated previously, there was no reduction in immunostaining at medium and large bundles in response to complement and thus complement inhibition could not further attenuate this. Single fibre and small bundle nodes of Ranvier also had significantly preserved ankyrin G and Caspr staining with eculizumab protection compared to ALXN3300 application (Fig. 4B and C; P < 0.001). Additionally, endogenous CFP was maintained in axons and bundles with eculizumab treatment compared to ALXN3300, essentially maintaining the normal overall architecture with a normal appearance (Fig. 4D). Functional protection of the distal nerve was also assessed. The sodium current as determined by perineural recordings in the distal nerve was preserved by application of eculizumab to anti-GD1a antibody and complement treated tissue and the results illustrated in Fig. 4D.
Calpain inhibition protects sodium channel and axonal protein integrity without preserving nerve currents
A consequence of membrane attack complex pore deposition in plasma membranes is the formation of a bi-directional, non-specific ion and water pore. At the nodes of Ranvier, the electrical function of the nodal axolemmal membrane is dependent on tightly regulated ion homeostasis and the consequences of this uncontrolled flux are likely to be considerable. To assess the consequence of the calcium component of ion influx, the protective effect of the synthetic calpain inhibitor AK295 was investigated. Neurofilament is a known calpain substrate (Chan and Mattson, 1999) and its protection by calpain inhibition in response to anti-ganglioside antibody-mediated complement-dependent injury at nerve terminals has been reported previously (O’Hanlon et al., 2003). In the present study, neurofilament at the neuromuscular junction was also significantly protected by 100 μm AK295 treatment compared to antibody and normal human serum treated, AK295 unprotected tissue (Fig. 5).
At the nodes of Ranvier in ex vivo whole-mount triangularis sterni muscle preparations exposed to antibody and normal human serum with and without AK295, assessments of Nav1.6, ankyrin G and Caspr immunostaining, and of perineural electrophysiological recordings were made. As expected, the extent of membrane attack complex deposition at nodes of Ranvier was completely unaffected by AK295 (data not shown). Nav1.6 immunostaining at single fibre and small bundle nodes of Ranvier was almost completely protected by AK295 treatment compared to unprotected treated muscle (Fig. 6A; P < 0.001). Ankyrin G and Caspr immunostaining was equally protected by calpain inhibition at single fibre and small bundle nodes of Ranvier (Fig. 6B and C; P < 0.001). For all nodal markers, the staining at nodes of Ranvier in medium and large bundles did not significantly differ as injury does not occur at these more proximal nodes of Ranvier.
To assess the protective properties of AK295 functionally, perineural recordings were conducted as previously described (Fig. 6D). Compared to normal control tissue currents, perineural Na+ and K+ currents were adversely affected by anti-GD1a antibody plus normal human serum, despite the presence of AK295. Thus, a similar loss of current flow to that seen in injured tissue as shown in Fig. 2D (lower panel) and Fig. 4D was observed. AK295 applied acutely to otherwise normal preparations had no neurotoxic effects (peak amplitude at time 0 and 20 min of 33.7 and 29.7 mV, respectively). These data indicate that calpain inhibition is able to prevent the destruction of major structural components at nodes of Ranvier, including Nav1.6 channels, but that despite this, loss of nodal conduction as assessed electrophysiologically still occurs.
Nodes of Ranvier in the nerve trunks are also vulnerable to anti-GD1a antibody and membrane attack complex-mediated calpain activation
Experiments on nerve trunks were conducted with anti-GD1a antibody and normal human serum as the complement source, in the presence and absence of calpain inhibition with AK295. A predominantly motor nerve (phrenic; 70% of myelinated fibres being motor, Langford and Schmidt, 1983) and purely sensory nerve (sural) were investigated in recording chambers during experimental incubations with normal human serum as the complement source, with continuous serial recordings followed by end-point immunohistology. Anti-GD1a antibody and complement (membrane attack complex) deposition were present at nodes of Ranvier in both phrenic and sural nerve; however their appearance was significantly different, being more elongated in distribution across the nodes of Ranvier in phrenic nerve compared to sural nerve (Fig. 7A and B; P < 0.001) . Furthermore, sural nerve nodes of Ranvier with antibody and membrane attack complex deposits, and yet intact Nav1.6 channel immunostaining, were often observed (Fig. 7C), a finding not seen in the phrenic nerve or its intramuscular branches. In terms of functional effects on CAP amplitudes, sural nerve CAPs remained stable or only modestly reduced over time (76.3 ± 9.6%; Fig. 7B), which was not significantly different from the controls.
In phrenic nerves, immunostaining of Nav1.6, ankyrin G and Caspr at nodes of Ranvier was quantified in response to anti-GD1a and normal human serum exposure. Having demonstrated nodal protein loss upon normal human serum exposure, experiments were also conducted in the presence and absence of AK295 (calpain inhibition) as for the ex vivo triangularis sterni preparation. Immunostaining of the extracellular domain of NrCAM, and moesin (a Schwann cell microvillal component) was also assessed, these being molecules within the nodal complex but predicted to be unaffected by calpain cleavage directly. There was a significant loss of immunostaining to Nav1.6, ankyrin G and Caspr in nerve exposed to antibody and complement, compared to control (Fig. 8A and C; P < 0.001). Unlike the reduction in Nav1.6, ankyrin G and Caspr staining, the NrCAM and moesin staining was retained but appeared mislocalized, being more diffusely spread throughout the node of Ranvier area in comparison with the staining pattern in control tissue (Fig. 8A and C).
Nav1.6 channel and Caspr staining was significantly protected by AK295 (Fig. 8B; P = 0.01 for both proteins), although this was more modest when compared with the levels of protection achieved at the distal nerve node of Ranvier. Protection of ankyrin G staining followed the same trend but did not achieve significance (P = 0.14). The retained but disrupted pattern of NrCAM and moesin immunostaining was not altered by AK295 treatment.
Under phase optics (differential interference contrast), a constant feature observed in phrenic nerve subjected to membrane attack complex deposition and injury was the swollen, granular appearance of the node of Ranvier, in comparison with control tissue, as visible in Fig. 8C. This subjective and unquantifiable appearance was unaffected by AK295 treatment (Fig. 8C, third and fourth column) but was consistently present. Two examples of each antibody staining pattern are shown for AK295 protected nodes of Ranvier (Fig. 8C, right). Extracellular recordings of phrenic nerve CAPs showed a large fall in amplitude over time (to 40.5 ± 10.7%) after treatment with anti-GD1a antibody and complement, in comparison with peak amplitude of the CAP prior to complement exposure (Fig. 8D and E). This fall in CAP amplitude was not significantly prevented by AK295 treatment (15.0 ± 8.7%, P = 0.3).
This study presents three major findings in relation to anti-GD1a ganglioside antibody-mediated acute neuropathy models. First, we demonstrate the increased vulnerability of very distal intramuscular nodes of Ranvier to antibody and complement mediated injury, in comparison with more proximal nodes that are relatively protected by the blood nerve barrier. Secondly, we show that axolemmal membrane attack complex pores at nodes of Ranvier result in calpain activation, which in turn causes (i) immunodetectable loss of key protein components of the nodal complex including Nav1.6 channels, most likely by protein cleavage leading to fragmentation (Iwata et al., 2003; von Reyn et al., 2009); and (ii) loss of function as demonstrated by the inability to record nerve terminal action potentials in distal axons. Thirdly, we show that electrical inexcitability of the nodes of Ranvier induced by membrane attack complex pores can occur in the presence of preserved gross structural integrity including that of key protein components (Nav1.6, ankyrin, Caspr), suggesting that failure of the axolemmal membrane to maintain ionic homeostasis when punctured by membrane attack complex pores is the critical factor in mediating axonal conduction block in this model.
The study has been facilitated by exploiting the intramuscular motor nerve node of Ranvier as a simple site relatively devoid of blood nerve barrier restrictions for analysing the pathological effects of locally or systemically delivered autoantibodies. Identifying the antibody access gradient allowed us to focus attention on the most vulnerable distal sites in intramuscular nerve bundles. We also demonstrated that any distal node of Ranvier effects did not result from concomitant latrotoxin-like, pre-synaptic injury that occurs in this model (Plomp and Willison, 2009). The preparation usefully permits dual experimentation on neuromuscular junctions and nodes of Ranvier, although the former site was not assessed in this study as it has been previously addressed (Goodfellow et al., 2005). The ex vivo focus of this study has allowed us to investigate very early pathological and electrophysiological features that result from membrane attack complex injury to nodes of Ranvier and their therapeutic responsiveness, events that would be untractable in man or animal studies in vivo.
The application of perineural recordings to monitor nodes of Ranvier electrophysiologically in our studies was essential as the motor end plate is concomitantly paralysed in this model, and endplate or muscle action potential measurement is thus not possible. The perineural recording technique allows for the recording of local electrical signals, resulting from the opening of ion channels from the pre-terminal, terminal and axonal regions of motor neurons (Mallart, 1985). When an electrode is inserted through the perineural sheath of a motor nerve close to nerve terminals, a waveform composed of two negative spikes can be recorded upon nerve stimulation. The first negative spike is attributed with inward Na+ current (INa, sensitive to tetrodotoxin) at the nodes of Ranvier in the axonal trunk, and the second negative spike represents the net local circuit current generated by the large outward current of K+ (IK) and a relatively small inward Ca2+ current at motor nerve terminals. The loss of recordable perineural currents in a distal-dominant pattern correlated with our immunohistological findings. At the distal node of Ranvier, the absence of recordable currents indicates a severe disruption of the ability of the node of Ranvier and motor nerve terminal to generate Na+ and K+ currents, respectively. This may either be due to calpain cleavage of the channels directly or due to the inability of the injured axon to maintain a resting membrane potential in the presence of membrane attack complex pores. The perineural current data obtained in the presence of calpain inhibition, in which channel integrity is preserved, indicate the latter mechanism is more likely. In the large intramuscular nerve bundles, which are relatively resistant to injury, the Na+ current was relatively preserved whereas the K+ current was reduced or absent, and interpreted as an inability to generate or propagate an action potential in the severely affected distal motor nerve that would be required to activate the terminal’s voltage dependent K+ channels (Braga et al., 1992).
Our previous studies have shown that the nerve terminal in this mouse model of anti-GD1a antibody-mediated acute motor axonal neuropathy is dependent upon membrane attack complex deposition (Goodfellow et al., 2005), and can be completely attenuated by the C5 neutralizing antibody, eculizumab (Halstead et al., 2008a) and other blockers of membrane attack complex formation (Halstead et al., 2008b). Here we also demonstrate the pivotal role for complement in mediating the disorganization of the nodes of Ranvier. This study does not directly rule out a possible contribution from C5a, whose formation is also blocked by eculizumab; however our previous studies on terminal axons have shown severe axonal injury is abrogated under C6 deficiency conditions (in which C5a is still formed), making a critical role for C5a unlikely. Furthermore, blockade earlier in the complement pathway, prior to C5, is also neuroprotective in our nerve terminal model (Halstead et al., 2005; Plomp and Willison, 2009). Nevertheless, the importance of other components of the complement pathway in peripheral nerve injury and repair should not be overlooked (Ramaglia et al., 2008). Our findings are supported by data from a chronic immunization model of acute motor axonal neuropathy in rabbits in which ventral root nodes of Ranvier are targeted by anti-GM1 antibody and complement that can be inhibited by nafamostat mesilate, although the precise mechanism(s) underlying this protection are likely to be different (Phongsisay et al., 2008). In the rabbit model, conduction block was attributed to lengthening of nodes of Ranvier accompanied by Nav1.6 channel loss or dispersion; whereas in our acute model, we conclude the conduction block arises principally as a result of node of Ranvier ionic imbalance due to axolemmal membrane attack complex pores. Irrespective of the precise mechanism, both studies raise the therapeutic prospect of using eculizumab in patients with acute motor axonal neuropathy and Guillain–Barré syndrome, as has been achieved in other membrane attack complex-mediated disorders (Hillmen et al., 2006). One advantage of the reliance of this mouse model on a heterologous (human) source of complement is that eculizumab is specific for human C5 and ineffective at neutralizing mouse C5. Therefore, therapeutic effects in a ‘humanized’ complement dependent injury using human therapeutic agents can be tested. It remains unknown why mouse complement cannot be sufficiently fixed in this model to cause pathological effects, in comparison with human complement, as previously discussed in detail (Willison et al., 2008).
The membrane attack complex pore comprises a transmembrane ∼5 nm pore allowing unselective, bidirectional flow of water, ions and soluble intracellular constituents of molecular weights up to 35 kDa (Podack et al., 1982; Simone and Henkart, 1982), including CFP (28 kDa). Thus in this model, the outward flow of CFP and its subsequent dilution in the extracellular environment is the most likely explanation for its disappearance, as it is not a calpain substrate and appears to be a very sensitive marker of pore formation. Even in the large intramuscular nerve bundles in which membrane attack complex is at undetectable levels, the CFP signal is attenuated, although this may alternatively be due to diffusion down axon with subsequent leakage in the more distally injured region.
At the nodes of Ranvier under physiological conditions, extracellular Ca2+ will flow intracellularly through membrane attack complex pores where one effect will be to activate calpain, as occurs at nerve terminals (O’Hanlon et al., 2003). Calpain-mediated proteolysis cleaves cytoskeletal and membrane proteins (Vosler et al., 2008), including sodium channels, as observed here. We used two Nav.1.6 antibodies that bind to different peptide domains on intracellular cytoplasmic loops between transmembrane channel subunits. Thus, the apparent ‘disappearance’ of Nav1.6 labelled by two antibodies observed here, over such a short timeframe, most probably equates to cleavage of the cytoplasmic loop(s), rather than more global disintegration, internalization or shedding of Nav1.6. Moreover, proteolysis of Nav1.6 intracellular loops may not significantly affect channel function since activation is preserved, the dominant effect being a failure of inactivation (von Reyn et al., 2009). Since ankyrin G links Nav1.6 to the cytoskeleton, it is also possible that the un-tethered Nav1.6 becomes mislocalized through lateral diffusion. The mislocalized but preserved immunostaining of Kv1.1 at the juxtaparanode, moesin in the Schwann cell microvilli and the extracellular domain of NrCAM at the node of Ranvier, alongside the phase optics images of the node of Ranvier, suggest that highly selective injury to the node of Ranvier axolemma was accompanied by local swelling and disorganization, with grossly preserved structural integrity of the axoglial unit over this timeframe. The mechanistic similarities between this model and the Nav1.6 loss at ventral root nodes of Ranvier reported in a chronic rabbit model of acute motor axonal neuropathy are unknown (Susuki et al., 2007b) .
Functional performance of the nodes of Ranvier under these injurious conditions was assessed with particular attention to Nav1.6, owing to its central role in nodal conduction. Injured nodes of Ranvier rapidly became electrically inactive, even when Nav1.6 and other calpain substrates were protected by AK295 treatment; indicating that failure to maintain ionic and water homeostasis owing to the presence of membrane attack complex pores, leading to membrane depolarization and inactivation of Nav1.6 channels, was the most likely mechanism, rather than Nav1.6 disruption. Ideally, ultrastructural examination of nodes of Ranvier would inform this; however electron micrographs of both control and affected nodes of Ranvier all showed fixation-related artefacts owing to the extended periods of time the nerve was maintained ex vivo, and were not suitable for analysis.
Our studies on phrenic and sural nerve trunks excluded the possibility that the vulnerability to anti-GD1a antibody-mediated injury was due to a distal to proximal GD1a antigen gradient in nerves (rather than a reflection of antibody access) and that concomitant, latrotoxin-like nerve terminal injury was responsible for any disruption to the juxtaterminal nodes of Ranvier. It also identified differential sensitivity of motor (phrenic) and sensory (sural) nerves to membrane attack complex-mediated injury that remains unexplained and may in part account for difficulties in interpreting this study in the context of previous work from us and others (Paparounas et al., 1999). Whether this quantitative or qualitative resistance of sural nerve to membrane attack complex-mediated injury provides insights into human acute motor axonal neuropathy, in which motor nerves are selectively affected, is unknown. In the currently used GD3s−/− mouse model, the sural nerve contains GD1a that is sufficiently available for antibody binding with complement activation, whereas in man, the levels of GD1a available for antibody targeting may be greater in motor than sensory nerve (De Angelis et al., 2001; Gong et al., 2002).
The above model describing very distal injury as a feature of acute motor axonal neuropathy corresponds well with existing clinical data, notwithstanding the co-occurrence in some cases of severe proximal injury (McKhann et al., 1991; Ho et al., 1997). In terms of underlying molecular mechanisms, the acute and severe motor nodes of Ranvier injury in this model may correspond to the initial phases of axonal conduction block seen in human acute motor axonal neuropathy, in which rapid onset but potentially reversible pathophysiology develops, prior to any cellular infiltration or axonal degeneration. The extent to which such events occur in man cannot be readily determined at a pathophysiological level as clinical and electrophysiological interrogation is very limited; however recovery in acute motor axonal neuropathy may be either very rapid and complete, or very prolonged with poor outcome, owing to extensive proximal axonal degeneration (McKhann et al., 1993; Hiraga et al., 2005b). The events at the nodes of Ranvier described here would correspond well with the early injury phase of a dichotomous outcome model (Gabriel, 2005). Critically, inhibition of the terminal complement product, membrane attack complex, as an early intervention seems essential from these data to limiting both acute injury and the development of more destructive long term pathology, while the inhibition of calpain activation downstream from membrane attack complex may also offer some partially additive benefit (Wang et al., 2004). Models addressing the role of eculizumab in more advanced stages of the pathological progression would also be useful. The expectation that clinical trials of complement or calpain inhibition in Guillain–Barré syndrome and its variants will inform this further is considerable.
The Wellcome Trust (#077041/Z/05/Z to H.J.W.). R.M. and K.N.G. were supported by the Medical Research Council doctoral training account awarded to the University of Glasgow.
Supplementary material is available at Brain online.
B6/Cg-TgN(Thy1-CFP) mice were kindly provided by Dr W. Thompson, Austin, TX, USA. AK295 was kindly provided by Dr J. Powers and J. Glass, Atlanta, Georgia. Dr Jaap Plomp is thanked for critical reading of the article.
compound action potential
cyan fluorescent protein
voltage-gated sodium channel
neuronal cell adhesion molecule