OUP user menu

Re‐expression of PSA‐NCAM by demyelinated axons: an inhibitor of remyelination in multiple sclerosis?

Perrine Charles , Richard Reynolds , Danielle Seilhean , Geneviève Rougon , Marie S. Aigrot , Adam Niezgoda , Bernard Zalc , Catherine Lubetzki
DOI: http://dx.doi.org/10.1093/brain/awf216 1972-1979 First published online: 1 September 2002


Multiple sclerosis is affecting ∼1 out of every 1000 individuals in the western world. After axons are denuded of myelin in the early stages of the disease, remyelination occurs, but eventually this process fails, and permanent disability is the result. During development, the polysialylated form of the neural cell adhesion molecule NCAM, PSA‐NCAM, is expressed at the axonal surface and acts as a negative regulator of myelination, presumably by preventing myelin‐forming cells from attaching to the axon. Removal of PSA‐NCAM from the axonal surface is a prerequisite for the initiation of myelination. We questioned whether, in multiple sclerosis, re‐expression of PSA‐NCAM by axons could occur, and therefore account for the failure of remyelination. Forty multiple sclerosis lesions from 24 different post‐mortem multiple sclerosis cases were selected by histological methods and analysed by immunohistochemistry. Demyelinated lesions and partially remyelinated lesions (shadow plaques) were studied. Controls consisted of post‐mortem brain tissue from patients with amyotrophic lateral sclerosis and without neurological disease. We showed that PSA‐NCAM, normally absent from adult brain, is re‐expressed on demyelinated axons in the plaques. Within shadow plaques, remyelinated axons do not express PSA‐NCAM. Re‐expression of PSA‐NCAM could act as an inhibitor of remyelination and participate in disease progression in multiple sclerosis.

  • Keywords: PSA‐NCAM; multiple sclerosis; remyelination
  • Abbreviations: GFAP = glial fibrillary acidic protein; PLP = proteolipid protein; PSA‐NCAM = polysialylated form of neural cell adhesion molecule


Multiple sclerosis is a common chronic inflammatory disease of the CNS characterized by multifocal demyelination. Spontaneous remyelination may occur, and remyelinated plaques, the so‐called shadow plaques, are areas of thin new myelin sheaths with short internodes, that allow restoration of a rapid, saltatory neuronal conduction. Shadow plaques may occur within acute plaques or at the edge of chronic plaques (Noseworthy et al., 2000; Wingerchuk et al., 2001). However, myelin repair is in most cases incomplete despite relative oligodendrocyte preservation within most plaques (Lucchinetti et al., 1999). Axonal pathology, which recently has attracted significant attention, also participates in this repair deficit (Losseff et al., 1996; Ferguson et al., 1997; Trapp et al., 1998; De Stefano et al., 1999; Evangelou et al., 2000a, b; Kornek et al., 2000; Lovas et al., 2000), but remyelination failure is not restricted to plaques with severe axonal loss. Even in areas where axonal preservation is found, no new myelin is detected. One possible explanation to account for the failure of remyelination in multiple sclerosis plaques would be the existence of mechanisms inhibitory to myelin repair.

We showed recently that myelination in the CNS was dependent on axonal signals, and that expression of the polysialylated form of the neural cell adhesion molecule NCAM, PSA‐NCAM, at the surface of axons acts as an inhibitor of myelination, presumably by preventing myelin‐forming cells from attaching to the axon. Disappearance of PSA‐NCAM from the axonal surface during development is coincident with the initiation of myelination. Furthermore, suppression of sialylated determinants by antibody‐mediated internalization, or enzymatic cleavage, increases myelination (Charles et al., 2000). Consistent with the idea that PSA‐NCAM ‘poises’ the axon in a pre‐myelination state, downregulation of PSA‐NCAM isoforms immediately precedes the onset of myelination. In the adult CNS, PSA‐NCAM is absent, except in those restricted areas of the brain which exhibit neurogenesis and/or plasticity (Seki and Arai, 1991, 1993; Rougon, 1993; Theodosis et al., 1999) such as the dentate gyrus, olfactory bulb and the hypothalamo‐hypophyseal system. It was therefore proposed that removal of PSA‐NCAM from the axonal surface is necessary to render the axon permissive for myelination (Charles et al., 2000).

We then questioned whether, in multiple sclerosis, demyelinated axons can recapitulate developmental events such as PSA‐NCAM re‐expression which might act as an inhibitor of remyelination. To test this hypothesis, we have analysed frozen post‐mortem brain tissue from 24 multiple sclerosis patients. Here we report that PSA‐NCAM is indeed re‐expressed on some demyelinated axons in the plaques, whereas it is absent from myelinated axons of the periplaque or normal‐appearing white matter. Moreover, we show that in shadow plaques, remyelinated axons do not express PSA‐NCAM. Our data provide evidence that PSA‐NCAM re‐expression on demyelinated axons may be a key molecular mechanism responsible for the partial failure of remyelination in multiple sclerosis.


Brain tissue collection

The study was performed on post‐mortem CNS tissue from 24 different multiple sclerosis cases. A total of 32 different tissue blocks were studied, with 40 lesions analysed. Among these, 30 consisted of plaque and periplaque, and 10 were identified as shadow plaques. In addition, two samples of normal‐appearing white matter were analysed. At least one block with plaque and periplaque was studied in each case. In four cases, several samples (2–5) from the same brain were studied. Material was collected in the Department of Neuropathology at the Salpêtrière Hospital in Paris (n = 9) and in the UK Multiple Sclerosis Tissue Bank at Imperial College School of Medicine in London (n = 15). Mean age was 59 years (range: 30–79) and mean disease duration was 26 years (range: 10–45). The course of multiple sclerosis was either secondary (nine cases), primary (six cases), progressive or relapsing–remitting (one case). In the eight remaining cases, the course of the disease was not specified. Controls consisted of post‐mortem brain tissue from patients with amyotophic lateral sclerosis (two cases) or without neurological diseases (three cases). Post‐mortem delay was 18 h (range: 4–49).

Neuropathology and immunohistochemistry

The brains were sliced coronally (1 cm thick), and alternate slices were either snap frozen or fixed. For frozen slices, 2 cm3 blocks were cut and immediately stored at –85°C. For fixed material, slices were heat‐sealed in a 4% paraformaldehyde (PFA)‐containing bag for 7 days, then 2 cm3 blocks were cryoprotected, frozen and stored at –85°C. Serial cryostat sections 10 µm thick were stained with haematoxylin–eosin (HE) and luxol fast blue (LFB)/periodic acid–Schiff (PAS) to assess inflammation, myelin loss and debris‐laden macrophages, respectively.

For immunohistochemistry of snap‐frozen material, sections were first fixed with 4% PFA in PBS (phosphate‐buffered saline) for 15 min, followed by 70% ethanol for 10 min, then incubated for 1 h in DMEM (Dulbecco’s modified Eagle’s medium) containing 50% horse serum and 10% fetal calf serum. Incubation with the primary antibody diluted in PBS containing 0.2% gelatin and 0.5% Triton X‐100 was carried out overnight at 4°C. After 1 h incubation in secondary antibodies at room temperature, samples were mounted in fluoromount‐G (Southern Biotechnology Associates, Birmingham, AL, USA) and examined by confocal microscopy (Omnichrome ion laser power supply on a Leica DRMB microscope). For immunolabelling of fixed material, the same method was used.

Selection of lesions

Plaques were defined as areas with histological evidence of demyelination confirmed on adjacent sections by the absence of proteolipid protein (PLP) staining. Periplaque white matter represented the strip of tissue of at least 5 mm adjacent to the border of the plaque. Normal‐appearing white matter was defined as an area, distant from the plaque, which showed no evidence of demyelination by macroscopic inspection and histology. Remyelinated shadow plaques were characterized by myelin pallor with LFB staining, and confirmed, in all cases, on adjacent sections by PLP staining showing abnormally thin myelin sheaths.


Mouse IgM monoclonal anti‐PSA‐NCAM antibody (hybridoma supernatant) was used diluted 1 : 4. This antibody has been characterized previously (Rougon et al., 1986; Häyrinen et al., 1995). The recognized epitope is a stretch of at least 10 neuramic acid residues α2–8 (Häyrinen et al, 1995). It gives no staining on tissue from mice lacking NCAM, the only characterized PSA carrier in mammals (Chazal et al., 2000). Rat IgG polyclonal anti‐PLP antibody (AA3 mAb), a gift from K. Ikenaka (University of Okazaki, Japan), was used diluted 1 : 10. Mouse IgG1 monoclonal anti‐neurofilaments (RMO‐55), a gift from V. Lee (University of Pennsylvania, Philadelphia, USA), was used diluted 1 : 10. Mouse IgG1 monoclonal anti‐phosphorylated neurofilaments (2F11; reacts with a phosphorylated epitope of a 70 kDa NF‐L, Dako, Trappes, France) was used diluted 1 : 500. Mouse IgG1 monoclonal anti‐phosphorylated neurofilaments (SMI 31; reacts with a phosphorylated epitope of a 200 kDa NF‐H and 150 kDa NF‐M, Interchim, Montluçon, France) was used diluted 1 : 500. Mouse IgG1 monoclonal anti‐dephosphorylated neurofilaments (SMI 32; reacts with a non‐phosphorylated epitope of NF‐H, Interchim) was used diluted 1 : 250. Rabbit polyclonal anti‐GFAP (glial acidic fibrillary protein) (Dako, Trappes, France) antibody was used diluted 1 : 4000. Fluorochrome‐conjugated rabbit antibodies against mouse IgG1 and IgM were from Tebu (Le Perray en Yvelynes, France) and were used diluted 1 : 200. Fluorochrome‐conjugated goat antibody against rat IgG was from Southern Biotechnology Associates and was used diluted 1 : 200. Fluorochrome‐conjugated sheep antibody against rabbit IgG was from Amrad Biotech (Boronia Victoria, Australia) and was used diluted 1 : 200.


Determination of the stage of the lesions

All plaques were classified as chronic by histological methods, characterized by a well‐demarcated region of myelin loss (Fig. 1). In only one case was there a persistent inflammatory component, as evidenced by PAS reaction (not shown). The 10 shadow plaques were all located at the edge of chronic plaques, and showed uniformly thin myelin sheaths. Astrogliosis, as assessed by anti‐GFAP staining, was detected in all plaques and was more pronounced in the plaque centre than in the periplaque zone. In two cases, axonal loss was prominent in the plaque. In the other cases, there was a relative preservation of axons within the demyelinated area, although swellings and ovoids, suggestive of axonal flow interruption and transections (Trapp et al., 1998), were often detected (Fig. 2A).

Fig. 1 PSA‐NCAM expression is detected within the multiple sclerosis plaque. Semi‐adjacent cryostat sections stained with LFB (A), anti‐PLP mAb (B), anti‐PLP and anti‐PSA‐NCAM antibodies (C). B is from the edge of a plaque and corresponds to the upper right corner of the lesion shown in A. (C) A magnification of the demyelinated area illustrated in the bottom left half of B. The demyelinated plaque appears as a well‐demarcated area of myelin loss, with LFB (A) and PLP staining (red) (B). In the plaque, the absence of PLP staining (red) contrasts with the presence of PSA‐NCAM‐positive linear structures (green) (C). (Scale bar: A, 500 µm; B, 300 µm; C, 40 µm).

Fig. 2 In multiple sclerosis plaques, PSA‐NCAM is expressed by axons. Immunostaining with anti‐PSA‐NCAM (A–D), RMO‐55 (A and D), 2F11 (B) and SMI‐32 mAbs (C) on cryostat sections from multiple sclerosis brain (A–C) and control brain from amyotrophic lateral sclerosis (D). In the multiple sclerosis plaque, PSA‐NCAM (green) is expressed on axons (stained with RMO‐55 antibody) (A). Not all axons are PSA‐NCAM positive: no PSA‐NCAM was detected on axons with highly phosphorylated neurofilaments (stained with 2F11 antibody) (B), while overlay with SMI‐32 immunostaining shows the co‐expression of PSA‐NCAM on axons with non‐phosphorylated neurofilaments (C). In controls (D), PSA‐NCAM (green) is not expressed by axons (red). Arrowheads in A indicate focal axonal swelling. Insets in A and C illustrate similar aspects in other multiple sclerosis lesions. (Scale bar A, B and D, 40 µm; C and insets, 20 µm).

PSA‐NCAM is re‐expressed by demyelinated axons

In 27 out of 30 plaques analysed, we detected positive PSA‐NCAM immunolabelling (Figs 1C and 2A–C) within the plaque, on long processes that were almost certainly axons. Axonal localization of PSA‐NCAM was demonstrated using RMO‐55 mAb, an antibody that specifically recognizes neurofilaments (Fig. 2A). Since none of the lesions analysed were located in areas known for persistent axonal PSA‐NCAM expression such as the dentate gyrus, hippocampus or hypothalamo‐hypophyseal system, this strongly suggests that PSA‐NCAM was re‐expressed on axons which had been demyelinated by the disease process. In three cases, no PSA‐NCAM was detected in the plaque. In two of these, axonal loss was severe. In the third case, there was relative axonal preservation. Interestingly, this lesion was the only one containing inflammatory infiltrates. Using antibodies recognizing different levels of neurofilament phosphorylation, we were then able to assess that PSA‐NCAM was not re‐expressed on axons with highly phosphorylated neurofilaments (stained with either SMI‐31 or 2F11 mAbs) (Fig. 2B). PSA‐NCAM expression was detected either on dephos phorylated neurofilaments containing axons (SMI‐32 positive axons) (Fig. 2C) or on axons with intermediate levels of neurofilament phosphorylation (RMO‐55 positive axons). In five lesion blocks, axons were sectioned longitudinally, allowing the quantification of PSA‐NCAM expressing axons. The mean percentage of PSA‐NCAM+/RMO‐55+ axons was 14% (range 11–19%). In contrast to the PSA‐NCAM re‐expression observed in the plaques, no expression of this molecule on axons was detected either in the periplaque (analysed in all cases) or in normal‐appearing white matter at a distance from the plaque (analysed in two cases). It should be noted that PSA‐NCAM expression was not limited to neuronal populations. A proportion of the PSA‐NCAM positive elements were double‐stained with GFAP antibodies; these were astrocyte cell bodies and processes within the plaque (Fig. 3A). In the five control cases, no PSA‐NCAM was detected on axons (Fig. 2D). Some PSA‐NCAM imunoreactivity was observed, however, on astrocytes (Fig. 3B).

Fig. 3 In multiple sclerosis plaques, PSA‐NCAM is also detected on astrocytes. Multiple sclerosis brain section double‐labelled with anti‐GFAP (red) and anti‐PSA‐NCAM (green) antibodies. Overlay of the two labellings illustrates the detection of PSA‐NCAM on GFAP positive astrocytes in the plaque (A) and in normal‐appearing white matter (B) (Scale bar A and B, 20 µm).

PSA‐NCAM is not expressed on remyelinated axons

Since axonal expression of PSA‐NCAM has been shown to be inhibitory to myelin formation, we speculated that its re‐expression on denuded axons could act as an inhibitor of remyelination. To gain further insight into this possibility, we analysed partially remyelinated lesions, the so‐called shadow plaques. Ten shadow plaques were examined. They were located on the edge of chronic lesions, and characterization by LFB staining and anti‐PLP immunolabelling demonstrated the co‐existence in the same plaque of denuded axons and remyelinated axons (Fig. 4A and B). In all shadow plaques examined, PSA‐NCAM immunoreactivity was detectable on astrocyte processes and cell bodies, and on axons (Fig. 4C). Interestingly, in these partially remyelinated plaques, among the PSA‐NCAM immunostained axons, a positive signal was never observed on the remyelinated internodes, but was restricted to demyelinated axons (inset Fig. 4C).

Fig. 4 Remyelinated axons do no express PSA‐NCAM. Myelin pallor detected with LFB illustrates a shadow plaque, at the edge of a chronic demyelinated plaque (A). B and C are from the shadow plaque area illustrated in A. (C) A magnification of the partially remyelinated area. Thin myelin sheaths are immunostained with an anti‐PLP mAb (red) (B and C). (WM = white matter, GM = grey matter). Remyelinated PLP positive internodes (red) do not express PSA‐NCAM (green) (C). Arrowheads indicate remyelinated axons, and the arrow points to a reactive astrocyte within the plaque. Inset: higher magnification showing two PSA‐NCAM+/PLP non‐myelinated axons, in the vicinity of two PSA‐NCAM/PLP+ myelinated axons (Scale bar A, 1 mm; B, 150 µm; C, 80 µm; inset, 40 µm).


The critical role of PSA‐NCAM in nervous system plasticity has been emphasized recently, but its precise role in nervous system development and repair has been difficult to ascertain directly, possibly because it has subtle, but nevertheless important functions.

PSA‐NCAM is involved in axonal pathfinding, nerve branching (Doherty et al., 1990; Zhang et al., 1992), cell migration (Wang et al., 1994, 1996), axonal fasciculation and synaptic plasticity. More recently, its role in CNS myelination has been reported (Charles et al., 2000). These data were obtained using the same anti‐PSA‐NCAM antibody as that used in the present study. It has been hypothesized that PSA‐NCAM, which is an extensively glycosylated molecule, might serve as a ‘repulsive strut’, keeping apposed membrane surfaces sufficiently apart until enough bona fide adhesion molecules are expressed appropriately on the neuronal surface to allow for a link up (Fannon and Colman, 1996). The enrichment of PSA‐NCAM on growth cone surfaces (Yamagata et al., 1995), and previous work on myelin (Charles et al., 2000) support the idea that this molecule plays a ‘delaying role’ in organizing a membrane–membrane adhesive link. In support of the idea that PSA‐NCAM essentially prepares a membrane surface to receive an adhesive input, the molecule, which normally disappears from the adult CNS, may be re‐expressed both on glial and neuronal cell surfaces in various pathological situations. After myelinotoxic lysolecithin‐induced lesions in the spinal cord, Schwann cells, oligodendrocytes progenitors and astrocytes expressing PSA‐NCAM have been detected (Oumesmar et al., 1995). After neuronal injury with either kainate or 6 OH‐DA (6 hydroxydopamine), PSA‐NCAM expression was shown on reactive astrocytes (Le Gal La Salle et al., 1992; Nomura et al., 2000). In the 6 OH‐DA model, re‐expression was also detected on a small number of dopaminergic neurones. Control of NCAM polysialylation in the adult CNS is regulated by enzymatic sialyltransferases (Breen et al., 1987) (named STX and PST activity). Therefore, re‐expression of PSA‐NCAM is probably explained by a reappearance of sialyltransferase activity in astrocytes and neurones. The regulation of sialyltransferase activity can occur at the transcriptional level. For example, it has been shown that in the hypothalamo‐hypophyseal system, a lactation‐induced decrease in PSA‐NCAM was associated with a significant reduction of STX and PST mRNAs (Soares et al., 2000). Alternatively, modulation of NCAM polysialylation could be related to a non‐transcriptional regulation of sialyltransferase activity. Evidence for a non‐transcriptional regulation of NCAM polysialylation has been reported by Bruses and Rutishauser (1998) who demonstrated that PSA regulation of expression in ciliary ganglion motor neurones was sensitive to calcium concentrations in intracellular compartments and independent of the level of sialyltransferase mRNA. Other modes of regulation of PSA expression are available to the cells. In addition to transcriptional and non‐transcriptional control of polysialyltransferases, neuronal electrical activity and nerve target interactions could also modulate PSA regulation of expression at the cell surface. These can, for example, induce differential delivery of pre‐synthesized PSA‐NCAM to the cell surface from intra‐cytoplasmic stores (Muller, 1992; Kiss et al., 1994; Wang et al., 1996) or, on the contrary, rapid endocytosis (Bouzioukh et al., 2001). In addition, PSA‐NCAM expression could depend on intrinsic neuronal properties (Aubert et al, 1998).

The mechanisms involved in PSA‐NCAM re‐expression on demyelinated axons are unknown, and several of them might co‐exist. The spatially restricted redistribution on demyelinated axons would be supported by a regulated exocytosis of PSA moieties at the surface of the demyelinated axon from intracellular stores not excluding an upregulation of polysialyltransferase activity in the concerned neurones. This regulated distribution could be related to the activation of calcium ion channels, which have been shown to be expressed on demyelinated axons (Kornek et al., 2001). Interestingly, this axonal expression of calcium channels is limited to the plaques and is absent from the normal‐appearing white matter. This regional distribution correlates with the restricted expression of PSA‐NCAM on the demyelinated portion of axons. Re‐expression of PSA‐NCAM and calcium ion channels could reflect axonal dysfunction of demyelinated axons. Here we show that PSA‐NCAM expression is restricted to axons with intermediate or low levels of phosphorylation. Neurofilaments in healthy myelinated axons are heavily phosphorylated. In contrast, non‐phosphorylated neurofilaments are abundant in demyelinated axons of multiple sclerosis lesions (Trapp et al., 1998). Therefore, re‐expression of PSA‐NCAM at the surface of these denuded axons, which displayed alterations in neurofilament phosphorylation, could be viewed as a plasticity response, in an attempt at neuroprotection following the demyelinating injury. This re‐expression could be linked sequentially with a modification of calcium ion channels inducing, in turn, PSA‐NCAM expression, or part of a more general developmental programme reinitiated by neurones. Re‐expression of PSA‐NCAM, however, would act as a double‐edged sword, since the protective role of PSA‐NCAM possibly could render the axon non‐permissive to remyelination. Interestingly, we found that within shadow plaques, PSA‐NCAM was not expressed on remyelinated axons. We could speculate that either the non‐expression or the downregulation of PSA‐NCAM from the axonal surface has allowed these axons to become permissive to remyelination. PSA‐NCAM expression was not restricted to axons, and numerous reactive astrocytes in the plaque were expressing PSA‐NCAM. To what extent astrocytic PSA‐NCAM expression participates in remyelination impairment remains to be established.

In multiple sclerosis, there are attempts to remyelinate, but myelin repair is in most cases insufficient after a few years of disease evolution. This failure to remyelinate appears to occur despite relative preservation of oligodendrocytes and axons in some demyelinated areas (Lassmann et al., 1998; Lucchinetti et al., 1999). Knowing that axonal PSA‐NCAM needs to be downregulated during development for myelination to proceed, our finding that in multiple sclerosis PSA‐NCAM is re‐expressed by demyelinated axons and is absent from remyelinated internodes supports the hypothesis of the existence of inhibitory signals of remyelination, PSA‐NCAM being one of these inhibitors.


This work was supported by INSERM and grants from ARSEP (Association de Recherche sur la Sclérose En Plaques), Fondation pour la Recherche Médicale and Neuroscience Federative Institute of Pitié Salpêtrière. P.C. had an INSERM fellowship (poste d’accueil INSERM). The UK Multiple Sclerosis Tissue Bank is supported by the Multiple Sclerosis Society of Great Britain and Northern Ireland (grant No. 458/97).


View Abstract