OUP user menu

Anti-disialoside antibodies kill perisynaptic Schwann cells and damage motor nerve terminals via membrane attack complex in a murine model of neuropathy

Susan K. Halstead, Graham M. O'Hanlon, Peter D. Humphreys, Deborah B. Morrison, Bryan P. Morgan, Andrew J. Todd, Jaap J. Plomp, Hugh J. Willison
DOI: http://dx.doi.org/10.1093/brain/awh231 2109-2123 First published online: 2 August 2004


Anti-disialoside antibodies (Abs) that bind NeuAc(α2–8) NeuAc epitopes on GQ1b and related gangliosides are found in human autoimmune neuropathy sera and are considered to be pathogenic. In a model system in mice, one mechanism by which anti-disialoside Abs have been demonstrated to induce paralysis is through a complement dependent blocking effect on transmitter release at the neuromuscular junction, similar to the effects of α-latrotoxin. Although direct targeting of presynaptic neuronal membranes occurs in this model, concomitant injury to perisynaptic Schwann cells (pSC) could indirectly contribute to this paralytic effect by influencing nerve terminal function and survival. To examine this possibility and the specific complement components that might mediate these effects, we exposed neuromuscular junctions in vivo and in vitro to an anti-disialoside Ab in conjunction with intact and selectively deficient complement sources. Using immuno-electron microscopy, we observed Ab deposits equally distributed on both neuronal and pSC membranes, and ultrastructural evidence of injury at both sites. Presynaptic neuronal injury was demonstrated functionally with microelectrode recordings and histologically as neurofilament loss. As hypothesized, concomitant pSC injury occurred, as indicated by abnormal uptake of ethidium dimer into pSC nuclei. The pSC and nerve terminal damage indicators correlated well with deposition of the pore-forming terminal complement component, membrane attack complex (MAC) in pSC and nerve terminal membranes. Furthermore, both neuronal and pSC injury were exacerbated in tissues from mice lacking the inhibitory complement regulator, CD59, where MAC formation is increased. These data demonstrate that both presynaptic neuronal membranes and pSCs are targets for anti-disialoside Abs, and that the injury to both sites is mediated by MAC and further regulated by CD59. This is the first demonstration that complement mediated pSC injury occurs in a model of autoimmune neuropathy and provides a rationale for investigating the possibility of pSC injury in equivalent conditions in man.

  • ganglioside
  • membrane attack complex
  • Miller Fisher syndrome
  • neuromuscular junction
  • perisynaptic Schwann cell
  • Ab = antibody
  • ACh = acetylcholine
  • BTx = α-bungarotoxin
  • C6def = C6 deficient
  • EM = electron microscopy
  • FDB = flexor digitorum brevis
  • immuno-EM = immuno-electron microscopy
  • mAb = monoclonal antibody
  • MAC = membrane attack complex
  • MEPP = miniature end-plate potential
  • MFS = Miller–Fisher syndrome
  • NF = neurofilament
  • NHS = normal human serum
  • NMJ = neuromuscular junction
  • pC6 = purified C6
  • pSC = perisynaptic Schwann cell
  • RT = room temperature


Anti-disialoside antibodies (Abs) that bind GQ1b and related disialylated gangliosides are serological markers for the Miller Fisher syndrome (MFS) variant of Guillain–Barré syndrome and other neuropathy subtypes (Willison et al., 2001; Willison and Yuki, 2002). Abundant evidence indicates that anti-disialoside Abs are important mediators of disease (Chiba et al., 1993; Kusunoki et al., 1996; Willison and O'Hanlon, 1999; Yuki, 2001; O'Hanlon et al., 2002). However, details of the sites of action and mechanisms by which the immunopathological events occur remain unclear (Sheikh and Griffin, 2001).

MFS is characterized by extraocular muscle paralysis, and this site-specific symptomatology can be accounted for by the relatively higher content of GQ1b in human extraocular nerves compared with spinal nerves (Chiba et al., 1993, 1997). In addition to their presence at nodes of Ranvier, complex gangliosides are enriched in presynaptic membranes, and some clinical evidence suggests that the neuromuscular junction (NMJ) may be affected in MFS (Uncini and Lugaresi, 1999; O'Hanlon et al., 2002; Wirguin et al., 2002). Experimental studies report a blocking action of anti-disialoside Abs at murine NMJs using the mouse hemi-diaphragm preparation (Roberts et al., 1994; Willison et al., 1996; Buchwald et al., 1998, 2002; Plomp et al., 1999; Bullens et al., 2000; O'Hanlon et al., 2001). By exposing mouse nerve-muscle preparations in vitro to a range of human and murine anti-GQ1b and anti-disialoside antisera and monoclonal Abs (mAbs), we have induced an α-latrotoxin-like, complement-dependent effect on neuroexocytosis of acetylcholine (ACh) at NMJs, followed by block of synaptic transmission and paralysis (Plomp et al., 1999; Goodyear et al., 1999; Bullens et al., 2000; O'Hanlon et al., 2001). Mice lacking complex gangliosides are resistant to these α-latrotoxin-like effects of anti-GQ1b Abs, indicating that expression levels of GQ1b determine susceptibility (Bullens et al., 2002).

The NMJ has a highly integrated anatomical arrangement comprising specialized domains in muscle, nerve and perisynaptic Schwann cells (pSCs), the latter supporting the underlying nerve terminal and modulating synaptic transmission (Castonguay et al., 2001; Rochon et al., 2001; Auld and Robitaille, 2003). Preliminary studies have suggested that pSCs accumulate anti-disialoside Ab and complement deposits in experimental nerve-muscle preparations, in addition to deposits on presynaptic membranes (O'Hanlon et al., 2002).

Complement can be activated via the classical, alternative and lectin pathways, and culminates in the incorporation of the lytic membrane attack complex (MAC; C5b-9) into target membranes. Although the action of anti-disialoside Abs at the mouse NMJ is clearly complement dependent, the precise pathway of complement activation is unclear (Plomp et al., 1999). In this study we have sought to identify pSCs as a site of deposition of an anti-disialoside mAb using light microscopy and immuno-electron microscopy (immuno-EM), and define the major complement factors and pathways involved in mAb-dependent NMJ injury. We show that the classical complement pathway activation with MAC formation is dominant and results in severe neuronal and pSC membrane injury, and pSC death.

Material and methods


The following mice were used: male Balb/c mice (Harlan, UK), male CD59−/− mice and wild-type CD59+/+ littermates, age 20–24 weeks on a mixed 129/Sv-C57Bl/6 background (Holt et al., 2001); male C6-deficient Peru-Coppock mice, age 10 weeks, on a C3H/He background (Orren et al., 1989); and ganglioside-deficient GalNAcT−/−/GD3s−/− double knockout (KO) mice and wild-type controls, age 4–6 weeks, on a C57Bl/6 background (Takamiya et al., 1996; Okada et al., 2002).

Abs and sera

The IgM anti-ganglioside mAb CGM3 was derived from mice inoculated with Campylobacter jejuni lipooligosaccharides bearing GT1a-like motifs. CGM3 reacts with the gangliosides GQ1b, GD3 and GT1a, which share a common terminal disialoside epitope, disialylgalactose [NeuAc(α2–8)NeuAc(α2–3)Gal–] (Goodyear et al., 1999). The mAb 22/18, a mouse IgM reactive with an irrelevant newt-specific carbohydrate, was used as a negative control as described previously (Kintner and Brockes, 1985; O'Hanlon et al., 2001). mAbs were quantified by enzyme-linked immunosorbent assay (ELISA) (Bethyl Laboratories, TX, USA). Normal human serum (NHS) from a single donor was freshly frozen and stored at −70°C (Plomp et al., 1999). C6- and C7-deficient (C6def, C7def) sera were obtained from patients with inherited total deficiency of C6 or C7. Purified C6 was obtained from Sigma (C3285; Poole, UK). Purified C7 (pC7) from human serum was obtained from Quidel (CA, USA). Both C6 and C7 deficiency prevent the complement pathway from progressing to MAC formation. All mAbs and sera were diluted and dialysed in Ringer prior to use.

In vitro muscle preparations

Mice were killed by CO2 inhalation, subject to Dutch and UK Home Office guidelines (Leiden DEC# 01055, UK PPL60/2305). Hemi-diaphragms and flexor digitorum brevis (FDB) muscles were dissected out onto Sylgard (Dow Corning, MI, USA) in a dish containing Ringer solution (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) at room temperature (RT) (20–22°C), pre-gassed with 95% O2/5% CO2. Unless otherwise stated, all in vitro bioassay preparations were exposed to the mouse anti-GQ1b IgM mAb CGM3 (50 µg/ml) for 2.5 h at 32°C, 30 min at 4°C and 10 min at RT, before being rinsed and then exposed to 40% NHS for 1 h at RT as described previously (Goodyear et al., 1999).

In vitro electrophysiological analyses of NMJ function

Intracellular recordings of miniature end-plate potentials (MEPPs), the small postsynaptic depolarizations resulting from spontaneous presynaptic release of single quanta of ACh packaged in a single vesicle, were made at hemi-diaphragm NMJs using standard recording equipment at 20–22°C. Randomly within the preparation, muscle fibres were impaled near the NMJ with a 10–20 MΩ glass micro-electrode filled with 3 M KCl. MEPP 0–100% rise times of <3 ms confirmed that the micro-electrode had been impaled close to the NMJ. Signals were digitized and stored for later off-line analysis.


Tissue was prepared as reported previously (O'Hanlon et al., 2001). To localize NMJs, Texas Red- and bodipy-labelled α-bungarotoxin (BTx, diluted 1/750 to 1.3 µg/ml; Molecular Probes, Eugene, OR, USA) were used. IgM was detected with TRITC-labelled goat anti-mouse IgM (1/300; Southern Biotechnology Associates, Birmingham, AL, USA). Complement components were detected by incubation for 1 h at 4°C with fluoroisothiocyanate (FITC)-labelled rabbit anti-C1q (diluted 1/100), FITC labelled rabbit anti-C3c (1/300) and mouse anti-human C5b-9 (1/50; all from Dako, Ely, UK) and rabbit anti-C4 (diluted 1/300; Sigma). Where required, incubation with secondary Ab (1/300; Southern Biotechnology Associates) was performed for 1 h at 4°C. Neurofilament (NF) staining using mouse mAb 1217 (1/750; clone SMI 32 reactive with non-phosphorylated NF; Affiniti Research Products Ltd, Exeter, UK) followed by goat anti-mouse IgG was performed as described previously (O'Hanlon et al., 2001). S100 staining was used to label pSCs in tissue fixed with 0.1% formaldehyde (VWR International, Poole, UK) then incubated with 0.1% Triton X-100 (VWR International) in phosphate-buffered saline (PBS) for 45 min at RT, using rabbit anti-cow S100 (diluted 1/200; Dako), followed by goat anti-rabbit-FITC.

Alternative and classical complement pathway activation

The alternative pathway of complement activation is Ca2+ independent, whereas the classical pathway requires Ca2+ at an early stage. To distinguish these, pairs of FDB muscles from Balb/c mice were incubated with CGM3 in Ca2+ free Ringer containing 4 mM EGTA (Sigma), rinsed three times in Ca2+-free Ringer, and then incubated with NHS in Ca2+-free Ringer. One FDB was transferred into NHS with normal Ringer, whereas the other remained in NHS with Ca2+-free Ringer. After rinsing in normal or Ca2+-free Ringer, tissue was processed for analysis of IgM, complement C4 and MAC deposition, and neurofilament loss. Experiments were performed in triplicate.

Physiological and immunohistological experiments using C6def mice and C6def human serum

To examine a role for MAC, total C6 deficiency was achieved in both the murine tissue and the exogenous complement source to prevent the formation of MAC but not affect earlier components of complement activation. Hemi-diaphragm preparations from C6def mice were exposed to CGM3 as described above. MEPPs were recorded before and after incubation. The ventral quarter of each hemi-diaphragm was then removed for baseline immunohistology. One preparation was then incubated with 40% C6def serum and the other with 40% NHS for 1 h at RT. Further electrophysiological parameters were then recorded. The dorsal quarter of each hemi-diaphragm was then removed for immunohistology. C6def serum reconstituted with pC6 (65 µg/ml) or NHS was then added to the remaining mid-sections of the hemi-diaphragm for a further 1 h at RT. Following electrophysiological recording, tissue was rinsed in Ringer and processed for immunohistology of IgM, C3c, MAC and NF. Experiments were performed in duplicate.

CD59−/− and CD59+/+ nerve-muscle preparations

Standard in vitro preparations of FDB muscles from male CD59+/+ and CD59−/− mice were exposed to CGM3 and NHS as above, then processed for immunohistology of IgM, C3c, MAC and NF. Experiments were performed in triplicate.

Measurement of pSC viability using ethidium homodimer

Ethidium homodimer 1 (EthD-1) is a membrane impermeant dye that labels with red fluorescence the nucleic acids of membrane-permeabilized cells (Molecular Probes). A 2 mM stock solution of EthD-1 (stored at −20°C) was diluted to 2 µM in oxygenated Ringer. Hemi-diaphragm and FDB preparations from Balb/c, CD59+/+ and CD59−/− mice were exposed to CGM3 and NHS, rinsed in Ringer and the bath volume was replaced with Ringer containing 2 µM EthD-1. The tissue was incubated in the dark at RT for 1 h, fixed in the dark in 0.1% formaldehyde for 1 h, rinsed in Ringer and frozen for immunohistology. NMJs were identified in 15 µm cryostat sections by staining with bodipy-BTx, and the frequency distribution of EthD-1-positive nuclei at end-plates was calculated.

Demonstration that the pSC lesion is MAC dependent

C7def serum was diluted to 30% in Ringer. In some instances pC7 was added to the C7def serum at physiological levels (56 µg/ml) and diluted to 40% in Ringer. Quarter diaphragms were incubated with Ringer or CGM3 as described above. The Ringer-treated preparation was then incubated with NHS for 1 h at RT, and the other three preparations with C7def serum, C7def serum + pC7 or NHS. Samples were then rinsed in Ringer, and quantified for EthD-1-positive nuclei as detailed above. The experiment was performed three times, and the data from each were pooled.

Passive immunization of mice with anti-GQ1b Ab

To demonstrate that any in vitro findings also occurred in vivo, passive immunization studies were conducted in male CD59+/+, CD59−/− and GalNAcT−/−/GD3s−/− double KO mice injected intraperitoneally with 1.5 mg (total dose) of CGM3, or an equivalent volume of PBS, followed 16 h later by NHS (0.5 ml). Animals were observed for 3 h following dosing with NHS, then killed by CO2 inhalation prior to tissue removal for immunohistology. Effective passive immunization was assessed by quantifying IgM, complement deposition and NF loss over the nerve terminal. In total, four pairs of CD59+/+and CD59−/− animals were treated with CGM3 and three pairs with PBS. One pair of CD59−/− animals, injected with either CGM3 or PBS, was perfused for ultrastructural analysis with PBS followed by 2% formaldehyde, 2.5% glutaraldehyde in PBS. In other instances, diaphragms from three CGM3- and two PBS-treated CD59−/− animals were assessed for pSC integrity with EthD-1 as detailed above.

Image acquisition and analysis

Digital images were captured by a Zeiss Pascal confocal microscope. Image analysis measurements were made using Scion Image analysis software (Scion Corporation Frederick, MD, USA). For quantitative analysis of IgM, C3c, C4, MAC and NF, three staining runs of each marker were performed on tissue from at least two individual hemi-diaphragm or FDB preparations, and quantified as described previously (O'Hanlon et al., 2001).

Electron microscopy and immuno-EM

For immuno-EM, duplicate experimental and control incubations were conducted on FDB muscle pairs from Balb/c mice. In vitro FDB preparations were incubated with CGM3 or the control mAb 22/18 (100 µg/ml). The samples were washed three times in fresh oxygenated Ringer, then fixed in situ with 2% formaldehyde and 2.5% glutaraldehyde in PBS for 1 h. Samples were then rinsed repeatedly in PBS prior to 30 min incubation with 1% sodium borohydride (Sigma) in ‘double salt’ PBS (containing 0.3 M NaCl). After rinses every 10 min for 90 min, the muscles were left overnight in PBS. Each FDB was separated into three, and to aid reagent penetration the muscle fibres were teased apart. Samples were then incubated with agitation overnight with goat anti-mouse IgM/1 nM gold conjugate (diluted 1/150; British Biocell International, Cardiff, UK), then treated with the IntenSE M silver enhancement kit (Amersham Biosciences, Little Chalfont, UK) for 15 min prior to rinsing in distilled water and embedding (O'Hanlon et al., 2001). Sets of reconstructed images of nerve terminal and pSC profiles from 21 NMJs (12 exposed to CGM3 and nine to control mAb 22/18) were analysed for the number of gold deposits per micrometre of membrane length of pSC, presynaptic and postsynaptic membranes using image analysis software (Image-Pro Plus 4.1; Media Cybernetics, Silver Spring, MD, USA). Gold deposits were counted directly using 10 000× and 30 000× magnified images of NMJs. Aggregates of two or more silver enhanced gold particles were counted as a single particle and the final count was calculated from the pooled results for the individual structures.

Ultrastructural analysis of NMJ from animals passively immunized with CGM3 or PBS was based on previous methods, with modification (O'Hanlon et al., 2001, 2003). NMJs were identified by the presence of postsynaptic junctional folds, and images recorded at 10 000× and 30 000× magnification were used for morphological assessment. Individual NMJs were composed of one or more separate nerve terminal profiles, and for each profile a series of measurements and counts were made. Profile perimeter, muscle contact length as a percentage of perimeter, and roundness, a measure of the profile's approximation to a circle, scoring 1 for a true circle to infinity for a straight line, were measured using Image-Pro Plus image analysis software (Media Cybernetics). The percentage of mitochondria within a nerve profile showing breakdown of internal structures was recorded for each terminal, as was the incidence of mitochondria touching the presynaptic membrane (scored 1 or 0). The presence of bundles of cytoskeletal fibres (scored 1 or 0) was assessed for each profile. The number of vesicles present in a 200 × 200 nm box placed over the presynaptic nerve terminal opposite the opening of a junctional fold gave an indication of the level of vesicle depletion. The numerical value presented for each profile was the mean of up to eight counts. The presence of pSC processes which separate a junctional fold from the nerve terminal (scored 1 or 0), and the incidence of pSC processes forming a ‘full wrap’ (scored 1 or 0) was assessed for each profile. The term ‘full wrap’ refers to encircling pSC processes that abut each other to give the appearance in the plane under observation of a complete separation of a nerve terminal profile from the underlying muscle (see Fig. 5 for examples).

Statistical analysis

Unless stated otherwise, statistical comparisons were made using a two-tailed Student's t-test employing a 5% level of significance.


Immunolocalization of anti-GQ1b Ab deposits at NMJs

Hemi-diaphragm preparations were exposed in vitro to the mAb, CGM3 and stained for IgM deposits in conjunction with BTx, or the Schwann cell marker S100. mAb deposits were clearly localized to end-plate regions being adjacent to and overlapping the BTx signal (Fig. 1A and B). At many NMJs, mAb deposits delineated one or more voids over the junctional area. Under phase-contrast optics, it was evident that these structures corresponded to cell nuclei (asterisks, Fig. 1B and C), and co-staining with S100 confirmed that these nuclear shadows belonged to pSC (asterisks, Fig. 1D–F), thus indicating that pSC membranes were being labelled by the mAb. The small overlap between the IgM deposits and the postsynaptic BTx signal observed in many micrographs suggested that neuronal membranes were also labelled with Ab deposits. In addition to the NMJ, CGM3 mAb deposits were seen within the blood vessel and capillary network, but were not observed within large nerve bundles or over muscle membranes, as identified by phase-contrast microscopy. Immuno-EM of the NMJ demonstrated that in CGM3-treated tissue, there was deposition of IgM on both the presynaptic neuronal and pSC membranes, but not over postsynaptic membranes, compared with control tissue (Fig. 2). Immuno-gold deposits were not present in significant amounts on muscle membranes or in intramuscular nerve bundles in the fields examined. Presynaptic neuronal membranes were labelled both in the synaptic cleft and over the membranes abutting pSCs. pSC membranes also appeared to be uniformly labelled, without evidence of polarization. In quantitative analysis, gold particles were equally distributed between neuronal and pSC membranes (Fig. 2E). These immuno-EM data are consistent with the confocal microscopic evidence that suggested both neuronal and pSC membranes accumulate CGM3 deposits.

Fig. 1

(AC) Localization of mAb CGM3 at mouse hemi-diaphragm NMJs. IgM deposits (green) are closely associated with end-plate regions (BTx; red). Areas of overlap appear yellow. IgM outlines pSC nuclear shadows (*) seen in phase-contrast (B and C). (DF) IgM deposits (red) also overlap (yellow) with the Schwann cell marker S100 (green); *pSC nuclei. (GJ) NMJs (BTx; red) exposed to mAb CGM3 plus NHS, immunostained for complement products (green). (G) C3c, *pSC nuclear shadows; (H) C1q; (I) C4; (J) MAC. (KM) EthD-1 (red) stained pSC nuclei at the NMJ (BTx or C3c, green). (K) One EthD-1-positive pSC. (L) Four NMJs with up to two EthD-1-positive nuclei (arrows). At NMJs labelled for C3c, complement deposits can be seen to surround the damaged pSC nuclei. (M) C3c deposits surrounding EthD-1-labelled pSC nuclei. (NQ) CD59−/− mice passively immunized with CGM3 or PBS, plus NHS. (N) CGM3 deposits (red) at NMJs (green). (O) MAC deposits (green) at NMJs (red). (P) NF (green) at control NMJs arborizes from the terminal axon into the junctional region. (Q) At CGM3-treated NMJs NF is absent. Scale bars: AJ, 10 µm; KQ, 40 µm.

Fig. 2

Immunogold localization of CGM3 deposits at mouse FDB NMJs. (A) No gold particles observed with control mAb 22/18. (BD) For CGM3-treated tissue, gold particles were abundant over the presynaptic nerve terminal (B and D) and also over pSC membranes in processes abutting the nerve terminal (C) and membranes around the cell body (D). Postsynaptic gold deposits were rarely seen. Randomly selected gold particles: asterisks in B and C; arrows in D. m = mitochondria; sv = synaptic vesicles; jf = junctional folds; psc = perisynaptic Schwann cells; nt = nerve terminal. Scale bar, 500 nm. (E) Quantitation of gold particles in CGM3 and control mAb 22/18-labelled tissue. Boxed numbers = total number of gold particles/total membrane length (µm) analysed. *Student's two-tailed t-test, significantly different from controls.

Nature and localization of complement deposits at NMJs

To determine the site of complement deposits and establish whether complement is being activated through the classical or alternative pathways, hemi-diaphragm and FDB preparations were exposed in vitro to the mAb CGM3 in conjunction with NHS as a source of complement. NMJs were then immunostained for deposits of C1q, C3c, C4 and MAC (Fig. 1G–J). Deposits of all complement components studied were readily visualized overlying NMJs, and the presence of C1q and C4 indicates that the classical complement activation pathway is in operation. The complement deposits directly overlayed the NMJ, and it was evident that the deposits abut the BTx staining, and also localize to pSCs, thus following an identical pattern to IgM deposits. Complement delineated voids, corresponding to the internal cytoplasm and nuclear regions of two pSC, are evident in Fig. 1G (asterisks). As described above, this pattern corresponds closely to IgM deposits, and in conjunction with the immuno-EM data indicates that both pSCs and the underlying nerve terminal accumulate complement activation products.

Classical and alternative pathway activation, and the role of MAC

The presence of classical pathway activation products, as demonstrated above, does not exclude the possibility of concurrent activation of the alternative pathway. The activation of C1q at the initiation of the classical pathway is Ca2+ dependent, as is the lectin pathway, whereas the alternative pathway runs to completion in the absence of Ca2+. Thus, to determine the relative involvement of classical and alternative pathways, FDB preparations were exposed in vitro to CGM3 and NHS as a source of complement in the presence of Ca2+-containing or Ca2+-free Ringer, then analysed for IgM, C4 and MAC deposits. Intra-terminal NF immunoreactivity was also assessed as an index of Ca2+-dependent calpain activation with subsequent NF degradation, as described previously (O'Hanlon et al., 2003) (Fig. 3A). As expected, deposition of IgM in preparations exposed to CGM3 and NHS was not effected in Ca2+-free Ringer, whereas C4 and MAC were absent. In the presence of Ca2+, C4 and MAC were detected at expected levels. These data indicate that alternative pathway activation is not the primary source of MAC deposition in this model, but do not exclude a possible role for MAC formation through the lectin pathway. NF signals were markedly reduced in Ca2+-containing Ringer, but were normal in Ca2+-free Ringer, as expected.

Fig. 3

(A) Complement activation by CGM3 plus NHS in Ringer with Ca2+ (classical pathway) and without Ca2+ (alternative pathway). CGM3 is equally deposited over NMJs under both conditions. C4 and MAC deposits (left axis) and NF loss (right axis) are only found in Ca2+-containing Ringer. *Student's two-tailed t-test, significantly different from Ca2+-free Ringer-treated samples. (B) Dependence of NF loss on MAC deposition, assessed in total C6 deficiency (C6def mouse with C6def human serum). No MAC is formed (left axis) and there is no NF loss (right axis). Reconstitution of C6def serum with pC6 leads to MAC deposits and NF loss. C3c is formed under both conditions (left axis). Student's two-tailed t-test: *significantly different from NHS treated samples; #significantly different from C6def serum-treated samples.

To investigate the role of MAC deposits, and a possible role for complement intermediates, including C3a and C5a, on the development of nerve terminal injury, nerve muscle preparations from C6def mice that were preincubated with CGM3 were exposed to human C6def serum. This double deficiency approach excluded the possibility of any endogenous (mouse tissue or serum derived) or exogenous (human serum derived) C6 from participating in the cascade, and should thereby completely prevent the formation of MAC. In the absence of C6, formation of C3c still occurred, but no MAC was detected and no nerve terminal injury was observed, as assessed electrophysiologically and by NF quantitation (Figs 3B and 4). Microelectrode recordings demonstrated the absence of any electrophysiological perturbation of NMJ function in the absence of C6 (Fig. 4A and B). Upon reconstitution of the C6def serum with pC6 (where MAC deposits could now be detected; Fig. 3B), high frequency MEPPs were observed, along with occurrence of fibre twitching (Fig. 4C), which is a previously described index of the α-latrotoxin-like effect (Jacobs et al., 2002). Paralysis of the preparation was tested by visual inspection of the muscle contraction following electrical stimulation of the phrenic nerve. Depletion of NF was also present under these conditions, indicative of neuronal injury at the nerve terminal (Fig. 3B).

Fig. 4

Electrophysiological analysis of NMJs under C6def conditions. Hemi-diaphragms of C6def mice (n = 2) were incubated serially in Ringer's medium, CGM3, C6def serum and C6def serum with added purified C6 (pC6). Spontaneous quantal ACh release at NMJs was measured as MEPPs with an intracellular microelectrode. Twitching muscle fibres (which occur when MEPP frequency at NMJs becomes very high) were visually scored under the microscope during the incubations as follows: 0 = no twitches; 1 = less than 10 fibres; 2 = moderate; 3 = extensive twitching observed within the preparation. MEPP frequency increased dramatically only when pC6 was added to the serum. (A) Typical examples of electrophysiological traces. (B) Average MEPP frequency (n = 2 muscles, 7–18 NMJs sampled per incubation). (C) Twitching of muscle fibres in preparations that were preincubated with CGM3 was observed only when C6 was present, i.e. in normal serum (upper panel) or C6def serum with added purified C6 (pC6), in the lower panel.

pSC injury monitored by electron microscopy

In view of the severe MAC-mediated injury observed in the neuronal elements of the NMJ, and the concurrent MAC deposition on pSC membranes, we looked for pSC injury that might be similarly mediated by MAC pore formation with subsequent intracellular Ca2+ ingress and osmotic swelling. Using electron microscopy to analyse NMJs in in vitro preparations, we observed both pSC and neuronal injury in CGM3- and NHS-treated preparations compared with control NMJs (Fig. 5). Control-treated NMJs were determined to have normal morphology: tightly packed synaptic vesicles, electron-dense mitochondria with well defined cristae and electron-dense pSC with no processes intervening into the synaptic cleft or nerve terminal (Fig. 5A and G). In CGM3 + NHS-treated tissue, nerve terminals were severely disrupted with electron lucent profiles, reduced synaptic vesicle density throughout the terminal and damaged mitochondria, as previously reported (O'Hanlon et al., 2001). With respect to pSCs, a spectrum of abnormalities was seen. One population of NMJs contain nerve profiles with newly formed pSC processes extending into the damaged nerve terminal (Fig. 5D–E). In some planes of section, pSC processes have entirely encircled an area of nerve terminal, completely separating it from the opposing postsynaptic area (‘full wrap’, see Material and methods for further description; Fig. 5E and F). This morphological appearance of process formation indicates pSC activation and therefore implies pSC viability at the time the processes were formed. In other micrographs, pSCs appear severely injured with electron lucent, sponge-like cytoplasm and damaged organelles. In these pSCs, nuclear disruption with swelling and membrane blebs is also evident, the overall appearance most likely representing necrotic cell death. A transitional phenotype was suggested in some micrographs, with the residual nerve terminal being penetrated by elongated pSC processes that themselves appeared to be undergoing degeneration. (Fig. 5H and I). This indicates that during the dynamic phase of evolution of the NMJ injury, pSCs may go through a phase of activation in response to nerve terminal injury, and subsequently (or contemporaneously) become targets for MAC-mediated lytic or sublytic injury.

Fig. 5

EM of nerve terminal injury induced by CGM3 in FDB muscle. (A) A normal NMJ profile exposed to control mAb 22/18 plus NHS. Synaptic vesicles are densely packed and mitochondria are healthy. pSC processes extend over the surface of the terminal, but do not extend into the synaptic cleft. (B) An NMJ exposed to CGM3 plus NHS. The presynaptic terminal is depleted of synaptic vesicles and mitochondria are damaged. pSC processes penetrate the synaptic cleft (arrowhead), separating the nerve terminal from junctional folds. (C) The nerve terminal is abnormally electron lucent. A pSC process is encroaching the synaptic cleft (arrowhead) with swollen, vesicle-laden processes, giving it a sponge-like appearance. (D and E) Nerve terminals exposed to CGM3 plus NHS. pSC processes are infiltrating the terminals (arrowhead). (E and F) pSC processes have formed a ‘full wrap’ (see Material and methods) around a portion of the nerve terminal. (G) A morphologically normal pSC nucleus (control mAb 22/18 plus NHS). (H and I) Structurally deranged pSCs (CGM3 plus NHS). pSC cytoplasm is swollen and electron lucent with plasma and nuclear membrane disruption. Presynaptic terminals are swollen, depleted of vesicles and contain damaged mitochondria. Postsynaptic structures are normal. m = mitochondria; jf = junctional folds; psc = perisynaptic Schwann cells; nt = nerve terminal; SV = synaptic vesicles. Scale bars for micrographs, 1 µm.

pSC injury monitored by a nuclear staining assay

In order to quantify the extent of this morphologically evident pSC injury, we used the DNA binding dye, EthD-1, as a rapid and easily quantifiable index of pSC plasma membrane integrity, and by extension, irreversible cell injury and death (Daly et al., 1992). In hemi-diaphragm preparations from Balb/c mice, we induced the CGM3 mAb and complement mediated lesion, and subsequently added EthD-1 to the organ bath and looked for labelling of the nuclei of injured and dying cells. In this model, clusters of one or more EthD-1-positive nuclei were observed at NMJs, overlaying or immediately adjacent to the BTx-labelled postsynaptic face (Fig. 1K and L). These EthD-1-positive nuclei were found to lie within complement delineated nuclear halos at the NMJ (Fig. 1 M), previously identified by S100 staining as pSCs. EthD-1-positive nuclei were not observed at NMJs from control tissues and no EthD-1 uptake was observed in the abundant muscle cell nuclei adjacent to the NMJ. Interestingly, and by way of control, a ribbon of nuclei in the degenerating muscle fibres along the cut edge of each hemi-diaphragm preparation did take up EthD-1, and thereby provided an indication that active dye had been applied to the preparation correctly.

NMJs were scored for the number of EthD-1-stained nuclei present. Out of a total of 575 NMJ sampled from three separate experiments on CGM3- and NHS-treated tissue, 270 (47%) had one or more EthD-1-positive nuclei. The frequency distribution of positive nuclei at NMJs in the sample is shown in Fig. 6A. In control studies, no positive nuclei were found overlying 196 NMJs exposed to Ringer and NHS. These data indicate that pSC membrane injury resulting in cell death is widespread over NMJs injured by CGM3 and complement. However, when MAC formation was prevented by using C7def serum instead of NHS, the incidence of EthD-1-positive nuclei was greatly reduced (Fig. 6C). In this instance, when using C7def serum, the low level injury to pSCs present in this sample is likely due to the presence of small amounts of endogenous C7 in the mouse tissues, allowing limited MAC formation to occur. This is supported by the observation that traces of MAC were evident in this sample (data not shown). When the deficiency was corrected by the addition of pC7, thus allowing the complement cascade to run to completion, the incidence of EthD-1-positive nuclei was restored to levels seen with intact NHS. These data indicate that, as with the nerve terminal injury, the pSC lesion is dependent on the deposition of MAC.

Fig. 6

pSC injury assessed by EthD-1 uptake. (A) The frequency distribution of EthD-1-positive pSCs at diaphragm NMJs from Balb/c mice exposed to CGM3 plus NHS (see also Fig. 1). (B) The frequency distribution of EthD-1-positive pSCs at FDB NMJs from CD59−/− and CD59+/+ mice exposed to CGM3 plus NHS. EthD-1-positive nuclei were more frequent in CD59−/− tissue (χ2-test, P < 0.001). (C) Analysis of MAC dependence of EthD-1 uptake into pSC nuclei in Balb/c mouse diaphragm incubated with CGM3 or Ringer, followed by either NHS, C7def serum, or C7def serum + pC7. In C7def conditions, EthD-1 uptake into pSC nuclei is greatly reduced, and can be restored to NHS levels by reconstitution with pC7. Boxed numbers = NMJs examined. *Significantly different from Ringer control; #significantly different from NHS treated tissue; χ2-test, P < 0.001.

Protection of NMJs from MAC-mediated injury by the complement regulator CD59

To determine the extent to which the complement regulatory protein CD59 might be attenuating the MAC-mediated nerve terminal injury, FDB preparations from CD59+/+ and CD59−/− mice were exposed to CGM3 in the presence of NHS and analysed for the deposition of IgM, C3c and MAC. When the pooled data from triplicate preparations was examined, the IgM and C3c levels in CD59−/− mice did not differ significantly from those detected in CD59+/+ mice [IgM (mean ± SEM): 103.1 ± 2.9%, P = 0.488; C3c: 100.8 ± 1.6%, P = 0.718]. However, MAC levels were almost doubled in CD59−/− mice compared with the CD59+/+ controls (180.0 ± 5.5%, P < 0.001). pSC death was monitored by EthD-1 uptake or exclusion, and was found to be significantly greater in CD59−/− mice (54% NMJ with one or more EthD-1-positive nuclei; n = 1956 NMJ) compared with CD59+/+ mice (28%; n = 2551; χ2-test, P < 0.001; Fig. 6B).

Passive immunization with anti-GQ1b Ab in vivo induces nerve terminal and pSC injury

To demonstrate that the in vitro findings described above could also be observed in vivo, CD59+/+ and CD59−/− mice were passively immunized with CGM3 plus NHS. Three hours after the addition of NHS, the CGM3-treated animals appeared ill and inactive compared with the PBS-treated controls. These mice were, however, still able to walk when stimulated, and further analysis is required to quantify any behavioural deficit. Diaphragms were analysed for deposits of IgM and MAC at the NMJ, and for NF degradation as an index of neuronal injury (Fig. 7). MAC deposits were increased nearly two-fold in CD59−/− mice. NF signals were reduced in both CD59+/+ and CD59−/− mice, but significantly more so in the latter (P = 0.02). Similar levels of IgM deposits were observed over NMJs in CD59+/+ and CD59−/− mice (Fig. 1N; P = 0.26), with no detectable IgM levels over the NMJs of animals injected with PBS. MAC levels were significantly lower in CD59+/+ compared with CD59−/− mice (Fig. 1O; P < 0.001) and virtually absent in PBS control-treated tissue. NF signals over the NMJ were greatly reduced in CD59−/− compared with CD59+/+ mice, in relation to their PBS treated controls (Fig. 1P and Q; P < 0.001).

Fig. 7

Analysis of IgM, MAC and NF levels at NMJs in CD59+/+ and CD59−/− mice exposed to CGM3 plus NHS. CGM3 deposits were unaffected, whereas MAC levels (left axis) were higher and NF levels (right axis) were lower in CD59−/− compared with CD59+/+ mice. Student's two-tailed t-test: *significantly different from respective PBS control; #significantly different from corresponding condition in CD59+/+ mice.

GalNAcT−/−/GD3s−/− double KO mice, which express neither GQ1b nor GD3 but only GM3, were found to be completely resistant to the effects of passively transferred CGM3 plus NHS in all of the above parameters measured [IgM (mean ± SEM): 6.8 ± 0.9%, P < 0.0001; MAC: 3.6 ± 0.9%, P < 0.0001; NF: 336.8 ± 14.8%, P < 0.0001; data expressed as percentage of wild-type control].

pSC injury as monitored by EthD-1 nuclear uptake was also found to be present in CGM3 passively transferred CD59−/− mice: 28% NMJ with one or more EthD-1-positive nuclei (n = 400 NMJ pooled from three different animals) compared with 0.8% in PBS-treated mice (n = 364 from two animals).

A sample of NMJ from one CD59−/− mouse passively immunized with CGM3 and NHS was compared ultrastructurally to NMJs from a comparable PBS-injected mouse and an existing normative database of normal controls from in vitro Balb/c hemi-diaphragm preparations. The CGM3-treated NMJs were found to exhibit many of the ultrastructural deficits previously noted in vitro (O'Hanlon et al., 2002, 2003). Within the terminal, synaptic vesicles were depleted and the incidence of damaged mitochondria was elevated (Fig. 8; Table 1). The nerve terminal profiles had a reduced level of contact with the postsynaptic muscle, and this was due, at least in part, to a greater number of profiles encircled by a ‘full wrap’ of pSC processes. The extent of pSC injury was not evaluated by either EthD-1 uptake or quantitative EM measurements in this pair of animals; however, the presence of well-formed pSC processes and the absence of widespread necrosis indicates a less severe pattern of pSC injury than observed in in vitro preparations. This is in accord with the reduced EthD-1 uptake by pSCs in the passive transfer model in comparison to the in vitro preparations (see above).

Fig. 8

Ultrastructure at NMJs in mice passively immunized with CGM3 plus NHS. (A and B) In PBS-exposed control tissue, mitochondria (m) appear normal, extensive reserves of synaptic vesicles (sv) are present adjacent to the presynaptic membrane, and where evident, cytoskeletal bundles (arrowhead) consist of regular parallel fibres. (C and D) NMJs from mice immunized with CGM3 show swollen mitochondria with distended or degenerating cristae, and reduced synaptic vesicle density. Cytoskeletal bundles are disorganized and indistinct (arrowhead). For quantitative analysis see Table 1. Scale bar, 500 nm.

View this table:
Table 1

Ultrastructural analysis of hemi-diaphragm nerve terminals in mice passively immunized with CGM3 or PBS

Combined control (102 NMJs, 257 profiles)PBS + NHS (24 NMJs, 98 profiles)CGM3 + NHS (18 NMJs, 59 profiles)
Vesicles per 200 × 200 nm box10.6 ± 0.212.2 ± 0.4*8.0 ± 0.5*
Damaged mitochondria (% of total)1.1 ± 0.31.8 ± 0.818.9 ± 4.4*
Mitochondria touching perisynaptic membrane (% of profiles)7.1 ± 1.78.9 ± 4.37.0 ± 3.9
Profile perimeter (µm)10.3 ± 0.59.7 + 0.610.6 ± 1.0
Profile roundness2.09 ± 0.061.79 + 0.091.62 ± 0.08*
Muscle contact (% of perimeter)45.9 ± 0.843.9 + 1.936.3 ± 2.5*
Cytoskeletal bundles present (% of profiles)35.4 ± 3.733.3 ± 6.730.5 ± 6.0
pSC processes ‘full wrap’ (% of profiles)1.2 ± 0.73.9 ± 2.718.6 ± 5.1*
  • Values are mean ± SEM. Combined control refers to normative data obtained by pooling data from Ringer treated controls preparations (see Material and methods).

  • * Significantly different from combined control;

  • significantly different from PBS control.


We here demonstrate both in vitro and in vivo that pSCs exposed to anti-disialoside Ab accumulate MAC deposits on their surface, resulting in lytic pSC death. Although the NMJ is involved in other Ab-mediated disorders (Lennon et al., 1995; Marvaud et al., 2002; McConville and Vincent, 2002; Lang and Vincent, 2003), pSCs have not previously been shown to be disease targets (Castonguay et al., 2001; Rochon et al., 2001; Auld and Robitaille, 2003). Our demonstration that pSCs can undergo autoimmune injury raises a new hypothetical mechanism that might in part contribute to the transient, distal nerve failure seen in the acute phase of some cases of MFS and Guillain–Barré syndrome (Ho et al., 1997; Uncini and Lugaresi, 1999; Wirguin et al., 2002; Spaans et al., 2003).

The immuno-EM data point strongly towards a presynaptic effect being the dominant site of the immunopathological effects of anti-GQ1b Abs at the NMJ, as suggested by our earlier studies (Willison et al., 1996; Goodyear et al., 1999; Plomp et al., 1999; Bullens et al., 2000; O'Hanlon et al., 2001, 2003; Jacobs et al., 2002). The anti-GQ1b mAb used in this study does not bind postsynaptic membranes, as reported using human MFS-associated anti-GQ1b antisera (Buchwald et al., 2001; Wessig et al., 2001). As an alternative to quantitative EM for assessing pSC injury, we established an easier method using EthD-1 nuclear uptake, which provides an indication of plasma membrane integrity (Daly et al., 1992). Here we have interpreted nuclear uptake of EthD-1 as an index of lytic cell death, whilst recognizing that some EthD-1 uptake might also occur in cells whose membrane injury is reversible (Jones and Morgan, 1991).

Using the EthD-1 assay, the proportion of MAC-damaged pSCs at any one end-plate in this model varies from none to all (Fig. 6). Since not all pSC nuclei are present in the plane of section, it is possible that the proportion of injured pSCs we observed is an under-estimate. However, the EM data indicate that some pSCs have been able to respond to nerve terminal injury by process formation, thereby indicating at least a proportion are viable. The extensive pSC injury seen here was unexpected, since in our previous report pSC process extension was observed without obvious necrotic injury (O'Hanlon et al., 2001). We have recently been able to attribute this to a striking difference in the susceptibility of pSCs to injury between previously used NIH Swiss and the current Balb/c strains of mice; this appears to be due to a difference in the level of CGM3 binding (G. O'Hanlon, unpublished data). We cannot attribute CGM3 binding to any specific ganglioside within the disialosyl series, but our evidence that the CGM3 mAb binding is directed at ganglioside epitopes rather than other disialylated structures has been reported previously (Bullens et al., 2002).

The pSC plays a key role in regulating the physiological function of the underlying nerve terminal (Castonguay et al., 2001; Rochon et al., 2001; Auld and Robitaille, 2003), and Schwann cells are also immunomodulatory (Armati and Pollard, 1996; Hartung et al., 1996; Gold et al., 1999; Wohlleben et al., 1999). Following pSC injury or death, it is likely that, until such time as the pSCs are replaced, denuded nerve terminals could be excessively vulnerable to transmission failure under inflammatory or stress conditions. Studies in the frog suggest that pSCs are not required for short-term maintenance of synaptic transmission, as we have also observed in mice (J. Plomp, unpublished data), but are essential for longer term nerve terminal maintenance (Reddy et al., 2003). The nerve terminal, including the pSC, lies outside the blood–nerve barrier, and is therefore vulnerable to Ab-mediated injury, in contrast to Schwann cells proximal to the terminal hemi-node, which are relatively protected by the blood–nerve barrier. Consistent with this, we have shown that CGM3 can bind nodes of Ranvier and myelinating Schwann cell cytoplasmic channels in histological and teased fibre preparations (Goodyear et al., 1999), yet in the physiologically intact preparations they are relatively protected from injury (G. O'Hanlon, unpublished data).

We recently observed that the presynaptic neuronal injury in this model arose in part through activation of the calcium-dependent protease, calpain, resulting in cleavage of nerve terminal NF (O'Hanlon et al., 2003). Both α-latrotoxin and MAC form pores in target membrane that allow the uncontrolled bi-directional passage of ions, small molecules and water (Acosta et al., 1996; Davletov et al., 1998; Newsholme et al., 1999; Ashton et al., 2000). It is thus likely that MAC insertion into presynaptic membrane leads to unregulated Ca2+ influx, and that the subsequent rise in intracellular [Ca2+] triggers the uncontrolled increase in MEPP frequency and concomitant calpain activation in the nerve terminal. The dependence of the pSC injury on MAC formation, as demonstrated by the absence of pSC injury under C7 deficient conditions, suggests that a similar pore-forming mechanism is responsible. Thus, MAC-mediated pore formation would also trigger pSC swelling and necrotic death, as seen.

CD59 is a potent complement regulatory protein that inhibits the formation of MAC, and is expressed on human and rat Schwann cells (Vedeler et al., 1994, 1999; Koski et al., 1996; Sawant-Mane et al., 1996). Our evidence indirectly suggests that CD59 is also present on pSC and nerve terminal membranes at the NMJ, in that it clearly plays a regulatory role in limiting pSC and nerve terminal injury. In experimental allergic neuritis, CD59 expression is increased on Schwann cells and is likely to protect them from MAC injury (Vedeler et al., 1999). Other complement regulatory proteins, including soluble complement receptor 1 (CR1/CD35), decay-accelerating factor (DAF, CD55) and membrane cofactor protein (CD46), have been identified on cultured Schwann cells (Koski et al., 1996), and, in the case of CR1, on myelinating Schwann cell membranes (Vedeler et al., 1999). In our studies, fresh human serum is used as heterologous source of complement as we have previously observed that mouse serum is incapable of providing sufficient activated complement to form detectable complement deposits in NMJ tissue (our unpublished data). This homologous restriction may be less pronounced when using mice deficient in DAF (which regulates the complement pathway through the accelerated breakdown of classical and alternative C3 and C5 convertases) or deficient in the combination of DAF and CD59. In this context, DAF deficiency has recently shown to facilitate induction of experimental myasthenia gravis in mice (Lin et al., 2002).

These studies provide the first evidence of anti-ganglioside Ab-dependent pSC injury, and indicate that the injury is mediated by MAC and regulated by CD59. The findings may have important implications for our understanding of the role for pSCs in modulating the NMJ and protecting it from injury, and the roles of complement products and regulators in the pathogenesis of murine models and human forms of autoimmune neuropathy. Our evidence would suggest that close attention be focused on electrophysiological and pathological studies of motor nerve terminals in human biopsy material, and that one therapeutic strategy to protect both neuronal and glial membranes might be through limiting the accumulation of MAC.


This work was supported by grants from the Wellcome Trust (H.J.W., G.M.O.H., P.D.H. and B.P.M.), Guillain–Barré Syndrome Support Group UK (S.K.H., H.J.W.), KNAW Van Leersumfonds (J.J.P.) and PPP Healthcare Medical Trust (D.B.M.).


View Abstract