OUP user menu

Preserved myelin integrity and reduced axonopathy in connexin32‐deficient mice lacking the recombination activating gene‐1

I. Kobsar, M. Berghoff, M. Samsam, C. Wessig, M. Mäurer, K. V. Toyka, R. Martini
DOI: http://dx.doi.org/10.1093/brain/awg072 804-813 First published online: 1 April 2003


Mice heterozygously deficient for myelin protein zero (P0) mimicking human Charcot–Marie–Tooth (CMT) disease 1B show T‐lymphocyte and macrophage upregulation in peripheral nerves, which aggravates and modulates the genetically mediated demyelinating neuropathy. In connexin32 (cx32)‐deficient (cx32def) mice, which mimic the X‐linked dominant form of CMT (CMTX), T‐lymphocyte and macrophage numbers are also significantly elevated in peripheral nerves. To test the hypothesis that immune cells are indeed pathogenic in this model, we cross‐bred cx32def mice with recombination activating gene‐1 (RAG‐1)‐deficient mice, which lack mature T‐ and B‐lymphocytes. In these immunoincompetent double mutants, the number of endoneurial macrophages was reduced. Furthermore, features indicative of myelin degeneration and axonopathic changes were mitigated in the RAG‐1‐deficient double mutants, whereas enlarged periaxonal Schwann cell collars, a hallmark specific for cx32‐mutants, were not reduced. Since both cx32‐ and P0 deficiency lead to similar immunopathogenic processes, we conclude that immune‐mediated demyelination may be a feature common to many CMT‐like neuropathies independent of the genetic origin.

  • Keywords: Charcot–Marie–Tooth neuropathy; connexin32; endoneurial macrophages; peripheral nervous system; T‐lymphocytes
  • Abbreviations: CMT = Charcot–Marie–Tooth; CMTX = X‐linked Charcot–Marie–Tooth; cx32 = connexin32; cx32def = connexin32 deficient; cx32wt = connexin32 wild type; nf = neurofilament; P0 = myelin protein zero; RAG‐1 = recombination activating gene‐1

Received June 11, 2002. Revised September 17, 2002. Second revision October 3, 2002. Accepted November 11, 2002


Demyelinating peripheral neuropathies are ultimately devastating disorders which lead to impaired function and degeneration of axons (Bjartmar et al., 1999; Martini, 2001; Young and Suter, 2001). A major cause for demyelination are glial mutations, which cause malformation or breakdown of the myelin sheath (Hanemann, 2001; Shy et al., 2001; Young and Suter, 2001). Autoimmune inflammation is another cause for demyelination in the PNS (Hartung et al., 1998c; Gold et al., 1999; Kiefer et al., 2001; Willison et al., 2002). We have previously shown that in mice heterozygously deficient for myelin protein zero (P0), both T‐lymphocytes and macrophages are increased in demyelinating nerves (Schmid et al., 2000; Carenini et al., 2001). Cross‐breeding of these mice with mutants either deficient in T‐lymphocytes or macrophage activation leads to a substantial alleviation of the neuropathy (Carenini et al., 2001; Schmid et al., 2000). Furthermore, reintroduction of normal immune cell progenitors into immunodeficient myelin mutants reverted the beneficial effect of immunodeficiency (Mäurer et al., 2001). These findings clearly reflect an involvement of immune cells in a Charcot–Marie–Tooth (CMT)‐like neuropathy that is primarily caused by a reduced gene dosage of the myelin‐related adhesion molecule P0.

Mice deficient in the gap junction component connexin32 (cx32) are myelin mutants mimicking another human hereditary neuropathy, the dominantly inherited form of X‐linked CMT (CMTX) (Anzini et al., 1997; Scherer et al., 1998). Although the mice do not show overt clinical abnormalities (Nelles et al., 1996), the pathological features of peripheral nerves are very similar to those seen in human biopsies of CMTX patients (Anzini et al., 1997; Martini, 1997; Scherer et al., 1998), indicating that the cx32‐deficient mice are an appropriate tool with which to study the corresponding pathomechanisms (Young and Suter, 2001). Unlike P0, which mediates myelin compaction (Martini and Schachner, 1997), cx32 is a component of channels that form a shortcut between the Schwann cell body and its most distal cytoplasmic extension, the periaxonal collar (Scherer et al., 1995; Balice‐Gordon et al., 1998). Blockade of this shortcut may result in a dramatic extension of this pathway by a factor of 103 in that ions and molecules have to travel a long way within the cytoplasmic turns of the Schmidt–Lanterman incisures and paranodal loops to reach the adaxonal compartment (Scherer et al., 1995). How this leads to a demyelinating neuropathy looking so similar to other CMT‐like neuropathies with different genetic origins is not yet known. Recent studies in the cx32 mutants suggested that immune cells might be implicated in demyelination, as has been shown for heterozygous P0 mutants (Kobsar et al., 2002). Particularly striking was the elevation of macrophages and the penetration of these cells into endoneurial tubes (Kobsar et al., 2002) resembling features of immune‐mediated demyelinating neuropathies, such as experimental autoimmune neuritis and the Guillain–Barré syndrome (Ballin and Thomas, 1969; Lampert, 1969; Ho et al., 1998).

In order to investigate whether immune cells are crucial determinants for demyelination in cx32‐deficient mice, we generated immunodeficient cx32 mutants by cross breeding the myelin mutants with mice deficient in the recombination activating gene‐1 (RAG‐1) lacking mature T‐ and B‐lymphocytes. We found that the double mutants not only displayed reduced numbers of endoneurial macrophages, but also showed a less severe myelinopathy and fewer axonopathic changes in comparison with cx32 mutants with an intact immune system.

Materials and methods


cx32‐deficient (cx32def) mice (cx32y/– or cx32–/– genotypes) (Nelles et al., 1996) and mice deficient in the RAG‐1 gene (RAG‐1–/–) (Mombaerts et al., 1992) have been generated by conventional knock‐out technology using homologous recombination in embryonic stem cells. For the initial step of generating cx32‐ and RAG‐1‐deficient double mutants, female cx32–/– mice were crossbred with RAG‐1–/– mutants (F1). cx32‐deficient mice were on a mixed C57/BL6 × 129sv genetic background, whereas RAG‐1 mutants were on a uniform C57/BL6 background. Male hemizygous cx32y/–/RAG‐1+/– and female heterozygous cx32+/–/RAG‐1+/– mice were then taken to generate the double mutants (F2). Thirteen‐month‐old hemizygous cx32y/–/RAG‐1–/– and homozygous cx32–/–/RAG‐1–/– mice were included in the study and were compared with cx32y/–/RAG‐1+/? and cx32–/–/RAG‐1+/? littermates. In addition, 13‐month‐old cx32 wild type (cx32wt) control mice (cx32y/+ or cx32+/+ genotypes) with a deficient (RAG‐1–/–) or intact (RAG‐1+/?) immune system were investigated. To avoid effects of the different genetic backgrounds, only littermates were analysed, as has been done in previous studies on the effect of immune cells on inherited demyelination (Schmid et al., 2000; Carenini et al., 2001; Mäurer et al., 2001). The breeding of the transgenic mice was approved by the Bavarian State authorities (Würzburg, Germany).

Determination of genotypes

Mice tail tips were taken and the genotyping of the control and mutant cx32 alleles were determined by conventional PCR using three specific primers (Moennikes et al., 1999). The cx32wt allele was detected using the oligonucleotides 5′‐CCATAAGTCAGGTGTAAAGGAGC‐3′ (primer 1) and 5′‐AGATAAGCTGCAGGGACCATAGG‐3′ (primer 2). The cx32 mutant allele was determined with primer 1 and the oligonucleotide 5′‐ATCATGCGAAACGATCCTCATCC‐3′ (primer 3).

The PCR was done in a final volume of 25 µl. The reaction product was denatured at 93°C for 2 min, followed by 40 cycles (93°C for 30 s, 66°C for 45 s and 72°C for 90 s), and a final extension at 72°C for 10 min. The PCR products were separated by gel electrophoresis and stained with ethidium bromide. The DNA alleles were represented as 550 bp (cx32wt) and 414 bp (cx32 mutant) fragments.

The RAG‐1 immune status was determined by FACS (fluorescence activated cell sorting) of peripheral blood cells stained with both CD4‐PE (phycoerythrin‐coupled anti‐CD4) and CD8‐FITC (fluorescein‐isothiocyanate‐coupled anti‐CD8) antibodies (PharMingen, Heidelberg, Germany) using a FACScan (Becton Dickinson, Heidelberg, Germany) (Schmid et al., 2000). Mice with CD4‐ and CD8‐positive T‐cells were classified as RAG‐1+/? animals, while mice lacking CD4‐ and CD8‐positive T‐cells were considered RAG‐1–/–.

Quantification of immune cells

Immunohistochemical analysis was performed on the femoral quadriceps nerve using 10‐µm thick serial cross‐sections from fresh frozen femoral nerves. Quantitative analysis of macrophages and CD8‐positive T‐cells was performed by horseradish peroxidase‐based immunohistochemistry (Carenini et al., 2001). For detection of endoneurial macrophages, antibodies to mouse F4/80 were used (1:300; Serotec, Eching, Germany); T‐cells were stained using antibodies to CD8 (1:1000; kindly provided by Professor R. Zinkernagel, Zürich, Switzerland). Sections of the spleen were taken as positive control for T‐cells when the CD8 antibody was administered. For negative control, the primary antibodies (F4/80 and CD8) were omitted. The number of macrophages and CD8‐positive T‐cells is given as cells per section. In addition, we counted the number of macrophages in ventral spinal roots of cx32def/RAG‐1+/? and cx32def/RAG‐1–/– mice by conventional electron microscopy (see below). Only endoneurial cells laden with myelin debris, devoid of a basement membrane and extending small, microvilli‐like processes, were considered as macrophages; their numbers were given as cells per section.

Tissue preservation for light and electron microscopy

Femoral nerves and lumbar (L2‐L4) ventral spinal roots were processed for light‐ and electron microscopy as reported recently (Carenini et al., 2001). The mice were transcardially perfused using 4% paraformaldehyde and 2% glutaraldehyde in 0.1 M cacodylate buffer. The nerves stayed in the same fixative overnight, followed by osmification and embedding in Spurr’s medium.

For light microscopic analysis, 0.5‐µm thick semithin sections from femoral nerves and ventral spinal roots were stained with alkaline methylene blue. Investigation of the pathological features was performed by blinded investigators (R.M. and K.V.T.) who were not aware of the RAG‐1 immune status. For electron microscopy, ultrathin sections (70 nm thick) were counterstained with lead citrate and investigated using a Zeiss EM 10B (Zeiss, Oberkochen, Germany).


The quantitative analysis of pathological changes was performed on ultrathin sections using a BioVision slow scan camera attached to the Zeiss EM 10B with the corresponding software analySIS 3.0 Doku (Münster, Germany). The following morphological parameters were assessed: (i) the number of myelinated axons with enlarged periaxonal Schwann cell collars (Anzini et al., 1997), demyelinated and thinly myelinated axons with a g‐ratio >0.9 (a reciprocal measure of myelin thickness; Friede, 1972); (ii) the number of fibres associated with supernumerary Schwann cells, which encompassed more than three‐quarters of the circumference of the corresponding fibre; (iii) the number of axonal regeneration units (Anzini et al., 1997); and (iv) the number of periaxonal vacuoles containing compressed, degenerating, or no axons. The g‐ratio was determined for 50 randomly selected nerve fibres of the femoral quadriceps nerve or ventral spinal roots. Since the numbers of myelin‐competent axons in femoral quadriceps nerves (500–600 axons) and in ventral spinal roots (700–1000 axons) differ, we provide the pathological parameters as percentages of the total number of myelin‐competent axons.

Neurofilament density was determined in 15–30 squares (0.04 µm2) of ultrathin sections of femoral nerves and ventral roots, of three individuals per genotype at a magnification of 25 000×. The confinements of the squares were randomly positioned on the sections and only squares that contained perpendicularly sectioned cytoskeletal elements were considered. Axons sectioned at the level of nodal regions or Schmidt–Lanterman incisures were not included.

For scoring the distribution of axon sizes, we measured the circumferences of 400–450 axons on ultrathin sections of femoral nerves (three individuals per genotype) at a magnification of 6400×.

Statistical analysis

Comparison of the pathological features of the different genotypes was done using the Mann–Whitney U‐test and statistical significance was defined for P < 0.05. All data are reported as means ± SD. To discriminate the distribution of axon sizes in femoral quadriceps nerves of the different genotypes, the Kolmogorov–Smirnov test was applied. Statistical analysis of the data was performed using MS Excel (Microsoft GmbH, Stuttgart, Germany) and SYSTAT (SPSS Inc., Chicago, USA). The graphs were constructed using SigmaPlot 2001 (SPSS Inc.).


In order to test the possibility that immune cells may play a role in the peripheral neuropathy of cx32def mice (cx32–/– and cx32y/– genotypes), we evaluated the number of CD8‐ and F4/80‐positive cells in femoral quadriceps nerves of cx32def and cx32wt mice. Indeed, CD8‐positive T‐cells and F4/80‐positive macrophages were significantly elevated in the cx32def mice (P < 0.05) (Fig. 1), with macrophages outnumbering T‐cells by a factor of ∼15.

Fig. 1 Quantification of (A) CD8‐ and (B) F4/80‐positive cells, and (C) representative immunohistochemistry with F4/80 antibodies on cryosections of femoral quadriceps nerves of cx32wt and cx32 mutant mice, either with or without RAG‐1 deficiency. (A) Elevated numbers of CD8‐positive cells can be seen in cx32def/RAG‐1+/? mice, whereas these cells are barely detectable in all other genotypes. Five mice of each genotype were investigated. Columns indicate mean values plus SD. (B) F4/80‐positive macrophages are strongly upregulated in cx32def/RAG‐1+/? mice. In comparison, in cx32def/RAG‐1–/– mice there are fewer macrophages than in cx32def/RAG‐1+/? mice. Mice with cx32wt expression show basic levels of resident macrophages irrespective of RAG‐1‐genotype. **Statistical significance between macrophage numbers in cx32def/RAG‐1+/? and cx32def/RAG‐1–/– mice (P < 0.01). Five mice of each genotype were investigated. Columns indicate mean values plus SD. (C) F4/80‐positive cells are rare in cx32wt mice (left), but substantially elevated in cx32def/RAG‐1+/? mice (middle). In addition, F4/80‐positive cells are more rounded and/or larger in comparison with macrophages of cx32wt mice, probably reflecting an activated or phagocytic state. In cx32def/RAG‐1–/– mice (right), the number of F4/80‐positive cells is clearly reduced in comparison with cx32def/RAG‐1+/? mice. Bars: 20 µm.

To characterize the possible implications of immune cells in cx32def mice, we cross‐bred the cx32def mutants with mice deficient in the RAG‐1 gene, which are devoid of mature T‐ and B‐lymphocytes (Mombaerts et al., 1992). In the immunodeficient cx32def double mutants, CD8‐positive cells were apparently absent from the peripheral nerves (Fig. 1). Moreover, a profound reduction of F4/80‐positive macrophages in the nerves of cx32def RAG‐1‐deficient mice was detectable, whereas RAG‐1‐deficiency had no influence on the numbers of macrophages in nerves of cx32‐positive mice (Fig. 1).

Numbers of macrophages were also determined in ventral spinal roots of cx32def/RAG‐1+/? mice and cx32def/RAG‐1–/– littermates by using conventional electron microscopy. Only cells laden with myelin debris, devoid of a basement membrane and extending small, microvilli‐like processes, were considered as macrophages. In ventral spinal roots of cx32def/RAG‐1+/? mice, 20.4 ± 4.4 macrophages per section were counted (n = 5), whereas ventral roots of cx32def/RAG‐1–/– littermates contained 7.8 ± 2.7 cells per section (n = 5; P < 0.001). Thus, although only a subpopulation of macrophages could be recognized by conventional electron microscopy, RAG‐1 deficiency led to a reduction of macrophages in ventral roots of cx32def mice.

Absence of the RAG‐1 gene leads to milder pathological changes in the femoral quadriceps nerves and ventral spinal roots of cx32def mice

First, we analysed the possible impact of RAG‐1 deficiency on pathological changes in cx32def mice at the light microscopic level. Semithin sections of femoral quadriceps nerves and ventral spinal roots from cx32def/RAG‐1+/? mice could clearly be discriminated from sections of cx32def/RAG‐1–/– mice due to the milder pathological features in the RAG‐1‐deficient myelin mutants (Fig. 3A and B).

We then quantified the pathological hallmarks typical for cx32def mice at the electron microscopic level (Fig. 2A). In femoral quadriceps nerves of cx32def/RAG‐1+/? mice, the typical pathological features of genuine cx32def single mutants were observed, such as enlarged periaxonal collars, demyelinated and thinly myelinated axons, and supernumerary Schwann cells. In addition, axonopathic changes, such as degenerating axons detaching from the inner loop and leaving empty large myelin sheaths, compressed or degenerating axons detaching from the inner myelin loop and leaving behind large periaxonal vacuoles, were detected. The number of enlarged periaxonal collars did not differ in cx32def/RAG‐1+/? and cx32def/RAG‐1–/– mice (Fig. 2A). In contrast, the number of demyelinated axons and the number of fibres associated with supernumerary Schwann cells in the form of onion bulbs was significantly lower in cx32def/RAG‐1–/– mice compared with cx32def/RAG‐1+/? littermates with an intact immune system (Fig. 2A). In addition, cx32def/RAG‐1–/– mice displayed significantly lower g‐ratios than cx32def/RAG‐1+/? mutants, reflecting thicker myelin sheaths in the immunodeficient myelin mutants (Fig. 2A). The number of axons associated with periaxonal vacuoles was also significantly reduced in cx32def/RAG‐1–/– compared with cx32def/RAG‐1+/? mice (Fig. 2A). The axonopathic phenotype of such vacuole‐associated axons is reflected by a substantially increased neurofilament (nf) density (517 ± 85 nf/µm2, n = 15) when compared with axons surrounded by apparently normal (251 ± 40 nf/µm2, n = 24), thin (346 ± 54 nf/µm2, n = 24) or no myelin sheaths (397 ± 38 nf/µm2, n = 30) (all P values <0.0001).

Fig. 2 Quantification of pathological features in femoral quadriceps nerves of (A) cx32def/RAG‐1+/? and cx32def/RAG‐1–/– mice, and (B) distribution of axonal circumferences of cx32def/RAG‐1+/?, cx32def/RAG‐1–/–, cx32wt/RAG‐1+/? and cx32wt/RAG‐1–/– mice. All parameters have been measured at the electron microscopic level. (A) Except for the enlarged periaxonal collars, all pathological key features are reduced in cx32def/RAG‐1–/– mice when compared with cx32def/RAG‐1+/? mice. Columns indicate mean values plus SD. Statistical significance with *P < 0.05 and **P < 0.01, respectively. Five mice of each genotype were investigated. (B) RAG‐1 deficiency in cx32def mice leads to a decrease in the share of smaller axons (5–15 µm circumference) and an increase in the share of middle‐sized axons (20–25 µm circumference) in comparison with cx32def/RAG‐1+/? mice (P < 0.0001). In addition, larger sized axons (40–50 µm circumference) are detectable in cx32def/RAG‐1–/– mice, whereas in cx32def/RAG‐1+/? mice the larger‐sized axons are completely absent. Note that there is no difference in axon size distribution between cx32wt/RAG‐1+/? and cx32wt/RAG‐1–/– mice. Three mice of each genotype were investigated.

Another feature indicative of axon abnormalities in cx32def/RAG‐1+/? mice was a substantial overall reduction of axon sizes compared with cx32wt/RAG‐1+/? mice (Fig. 2B). In general, the distribution of axon sizes in femoral quadriceps nerves of cx32def/RAG‐1–/– mice was intermediate when compared with cx32def/RAG‐1+/? and cx32wt mice, and these differences were highly significant (P < 0.00001, Kolmogorov–Smirnov test). Within the group of cx32 mutants, RAG‐1 deficiency led to an improved axonal phenotype in that there was a decrease in the share of smaller axons (5–15 µm circumference) and an increase in the share of middle‐sized axons (20–25 µm circumference) compared with cx32def/RAG‐1+/? mice (P < 0.00001). Also, larger sized axons (40–50 µm circumference) were detectable in cx32def/RAG‐1–/– mice, whereas in cx32def/RAG‐1+/? mice the larger‐sized axons were completely absent (Fig. 2B). Another abnormality was regenerating axon clusters. The numbers of regenerating axon clusters did not differ between both genotypes (not shown). However, the number of regenerating axons within these clusters was significantly higher in the immune‐deficient myelin mutants (2.57 ± 0.22) compared with the cx32def/RAG‐1+/? mice (2.20 ± 0.24; P < 0.03).

By investigating the ventral roots of cx32def/RAG‐1+/? mice, we generally found similar pathological alterations as in femoral quadriceps nerves. However, some pathological hallmarks differed in their frequency. For instance, the number of demyelinated axons was increased by a factor of ∼5, whereas the number of fibres associated with supernumerary Schwann cells was reduced. The pathological characteristics were even more strikingly mitigated by the immunodeficiency than in the femoral quadriceps nerves, so that ventral roots of cx32def/RAG‐1–/– mice displayed only very mild features indicative of degenerative processes (Fig. 3). This was reflected statistically by significant differences in the frequency of abnormal features (Fig. 4). Particularly striking was the profound reduction of axonopathic features in immunodeficient mice. In the immunocompetent cx32 mutants, myelin vacuoles containing axons with a similarly increased neurofilament density (528 ± 83 nf/µm2, n = 17), as in femoral nerves, were amply present, whereas these features were almost absent in the cx32def/RAG‐1–/– mice (Figs 3 and 4). Similar to the femoral quadriceps nerve, the number of enlarged periaxonal collars was not reduced in the ventral roots of immunodeficient mutants (Fig. 4).

Fig. 3 (A, B) Light and (C, D) electron microscopy of ventral spinal roots of cx32def/RAG‐1+/? (A, C) and cx32def/RAG‐1–/– (B, D) mice. Note prominent pathological features in cx32def/RAG‐1+/? mice, such as thin myelin, demyelinated axons (* in C and D) and periaxonal vacuoles (**). Such features are rarely seen in cx32def/RAG‐1–/– mice (B, D), and most of the myelinated profiles appear nearly normal. Enlarged periaxonal collars (arrows in C and D) are amply present in both genotypes. BV = blood vessels. Bars: 10 µm in (A and B), and 2 µm in (C and D).

Fig. 4 Quantification of pathological features in ventral spinal roots of cx32def/RAG‐1+/? and cx32def/RAG‐1–/– mice at the electron microscopic level. Except for the enlarged periaxonal collars, all pathological features investigated are reduced in cx32def/RAG‐1–/– mice when compared with cx32def/RAG‐1+/? mice. Columns indicate mean values plus SD. *(P < 0.05) and **(P < 0.01) indicate statistical significance. Five mice of each genotype were investigated.


We demonstrate that in peripheral nerves of ∼ 1‐year‐old cx32def mice, the number of CD8‐positive T‐lymphocytes and F4/80‐positive macrophages is significantly elevated when compared with wild‐type littermate mice. Generating an immunodeficient cx32def mutant by cross‐breeding the myelin mutants with RAG‐1‐deficient mice did not only lead to the expected absence of endoneurial lymphocytes, but also to a substantial reduction in the number of endoneurial macrophages in both femoral nerves and ventral roots. Typical pathological features of the cx32 mutant, such as thinly myelinated or demyelinated axons and supernumerary Schwann cells, were highly reduced in the immunocompromised double mutants. Also, axons associated with myelin vacuoles and with highly increased neurofilament density, a typical feature of myelin‐related axonopathy (de Waegh et al., 1992) or axonal atrophy (Yin et al., 1998), were much less frequently observed in immunocompromised cx32 mutants. In addition, axon calibres were generally less reduced in size in cx32def/RAG‐1–/– mice than in cx32def/RAG‐1+/? mice.

The present observations are similar to those found in P0+/–/RAG‐1–/– mice. This double mutant, which was on a uniform C57/BL6 background, showed a substantial amelioration of the pathological phenotype in comparison with P0+/–/RAG‐1+/? littermates (Schmid et al., 2000; Mäurer et al., 2001). In addition, deficiency in the α‐subunit of the T‐cell receptor or absence of macrophage colony‐stimulating factor led to an improved neuropathology in P0 mutants, although the genetic backgrounds were heterogeneous (Schmid et al., 2000; Carenini et al., 2001). These combined observations reflect that it is highly improbable that our present findings are the result of comparing littermates of a non‐uniform genetic background, and thus strongly suggest that in two different models of inherited demyelination, immune mechanisms are relevant to pathogenesis.

The amelioration of the pathological phenotype in cx32def/RAG‐1‐deficient mice was related to several types of degenerative features, such as demyelination, generation of supernumerary Schwann cells and axonopathic changes. This suggests that these processes are driven or fostered by activated T‐lymphocytes and macrophages. The enlarged periaxonal collars, a typical pathological hallmark of cx32def mice, were not the only features that could not be rescued by compromising the immune system. Thus, the development of this pathological feature may be independent of the immune system and might reflect the direct subcellular consequence of impaired cx32 function.

The reduction of axonopathic changes, which is typical in cx32‐deficient mice (Anzini et al., 1997; Scherer et al., 1998) and CMTX patients (Senderek et al., 1998, 1999; Hahn et al., 2001; Nakagawa et al., 2001), deserves particular consideration with regard to the involvement of the immune system. Since there is no hint of a generally reduced susceptibility of axons in RAG‐1‐deficient mice to neuropathic changes (Rambukkana et al., 2002), we consider the improved axon preservation a result of the immune deficiency. How inflammatory cells interfere with axons in cx32 mutants and possibly in CMTX patients is not known, but it is well established that in acquired immune‐mediated neuropathies, such as Guillain–Barré Syndrome and its animal model experimental autoimmune neuritis, neurotoxic compounds produced by inflammatory cells damage axons (Hartung et al., 1993, 1998a, 1998b; Gold et al., 1999; Mäurer et al., 2002b). As a seemingly contradictory feature, it has recently been suggested that immune cells may also exert neuroprotective rather than destructive effects in inflammatory processes of the CNS by the secretion of neurotrophic factors (Hauben et al., 2000; Hohlfeld et al., 2000). In our models of inherited neuropathies, i.e. in P0+/– and cx32def mice, we have not formally examined whether endoneurial immune cells express neurotrophic factors but, even if they do, these potentially beneficial effects obviously do not compensate for the axonopathic immune effects in cx32def mice.

What are the cellular and molecular events that lead to the immune‐mediated exaggeration of genetically determined neuropathy? In models of acquired demyelinating neuropathies, such as adoptive‐transfer experimental autoimmune neuritis in rats, activated neural‐specific CD4+ T‐lymphocytes of the helper type are the first T‐cells to invade peripheral nerves (Hartung et al., 1998a, b; Gold et al., 1999; Mäurer et al., 2002b). When these cells meet resident macrophages that present the appropriate antigen in the context of the class II major histocompatibility complex, a robust autoinflammatory reaction occurs due to the release of proinflammatory cytokines from the T‐helper cells, such as interleukin‐2, interferon‐γ and tumor necrosis factor‐α (Hartung et al., 1998a, b; Gold et al., 1999; Mäurer et al., 2002b). This causes upregulation of endothelial and lymphocytic adhesion molecules and results in an opening of the blood–nerve barrier, leading to further invasion and activation of T‐cells and cells of the monocyte/macrophage lineage. Activation of macrophages by the proinflammatory cytokines eventually leads to major detrimental effects by additional cytotoxic cytokines, free radicals, nitric oxide, prostaglandins and direct phagocytic attack of myelin. In addition, direct cytotoxicity of lymphocytes may be relevant (Hartung et al., 1993).

In our myelin mutants, similar cellular players are involved but there appear to be fundamental differences in comparison with non‐genetically based acquired neuropathies. In both P0 and cx32 mutants, elevated numbers of macrophages are found in the nerves before an elevation of T‐lymphocytes is observed (Schmid et al., 2000; Kobsar et al., 2002), suggesting that the initial elevation of macrophages is not the consequence of T‐cell infiltration. It is not known which factors mediate the increase and activation of the macrophages in the myelin mutants, but it is tempting to speculate that Schwann cell‐related factors are involved. Schwann cells may suffer from the direct consequences of cx32‐ or P0‐deficiency, as shown ultrastructurally by the swollen periaxonal collars or decompaction of myelin, respectively (Martini et al., 1995; Anzini et al., 1997). Such direct reactions may somehow induce unknown intracellular signals that cause the secretion of chemokines/cytokines by the Schwann cells, which may attract or activate macrophages. One possible candidate is macrophage colony stimulating factor, since its absence results in much lower numbers of macrophages in the mutant nerves (Carenini et al., 2001). In addition, chemokines (Huang et al., 2000), and in particular monocyte chemoattractant protein‐1, are of interest, since it is upregulated by Schwann cells of injured nerves (Toews et al., 1998; Taskinen and Roytta, 2000; Subang and Richardson, 2001; Tofaris et al., 2002). Similar to acquired neuropathies, the activated macrophages may then attract the CD8+ T‐cells that in turn may regulate the numbers of macrophages (Hartung et al., 1998a). A clear difference to the acquired neuropathies concerns the amount and types of T‐lymphocytes. In experimental autoimmune neuritis, there is a robust infiltration of CD4+ T‐lymphocytes, particularly when high levels of antigens are used for active immunization (Hartung et al., 1993). This is followed by a dominance of CD8+ cells during recovery (Mäurer et al., 2002b). In myelin mutants, CD4+ lymphocytes are not detectable in the nerves at any stage of the disorder (Schmid et al., 2000) and the number of CD8+ cells is relatively low. Since in both cx32 and P0 mutants, CD8+ T‐cells are outnumbered by macrophages by a factor of 15 or 20, respectively, we favour the idea that lymphocytes regulate macrophages rather than directly impair Schwann cells or axons (Mäurer et al., 2002a).

Our observations lead us to suggest that chronic low grade inflammatory mechanisms may be crucial to some of the hereditary demyelinating neuropathies. How the myelin‐related mutations lead to the activation of the immune system is not yet clear. According to our models, Schwann cells are good candidates for initiating an inflammatory cascade starting with an activation of macrophages that are further triggered by regulatory T‐cells. If this holds true, the molecular mechanisms leading from Schwann cell‐related mutations to macrophage recruitment may be the crucial issues that should be investigated, and, moreover, the intereaction between T‐lymphocytes and macrophages. In relation to inherited peripheral neuropathies in humans, the discovery of a role for immune‐mediated pathology in some forms of the disorders gives hope for an effective immunotherapeutic intervention in the corresponding cases. As a hypothesis, the exaggeration of neural lesions by inflammatory mechanisms might also be relevant for some other degenerative disorders of the peripheral and central nervous system, which should be the focus of future forms of neurological research.


We gratefully acknowledge Prof. Thomas Hünig, Prof. Ralf Gold and Dr Harald Hofstetter (University of Würzburg) for valuable discussions. We thank Dr Thomas Ott (University of Bonn) for providing us with breeding pairs of cx32‐deficient mice and for initial help with genotyping. We are particularly grateful to Heinrich Blazyca and Carolin Kiesel for excellent technical assistance, and Helga Brünner and Dr Bettina Holtmann for skillful animal care. This study was supported by the Deutsche Forschungsgemeinschaft (SFB 581 and Priority Program ‘Microglia’), the Gemeinnützige Hertie‐Stiftung and by University Research Funds from the State of Bavaria.


View Abstract