Axon–glial interaction in the CNS: what we have learned from mouse models of Pelizaeus–Merzbacher disease

Abstract

In the central nervous system (CNS) the majority of axons are surrounded by a myelin sheath, which is produced by oligodendrocytes. Myelin is a lipid-rich insulating material that facilitates the rapid conduction of electrical impulses along the myelinated nerve fibre. Proteolipid protein and its isoform DM20 constitute the most abundant protein component of CNS myelin. Mutations in the PLP1 gene encoding these myelin proteins cause Pelizaeus–Merzbacher disease and the related allelic disorder, spastic paraplegia type 2. Animal models of these diseases, particularly models lacking or overexpressing Plp1, have shed light on the interplay between axons and oligodendrocytes, and how one component influences the other.

Introduction

The central nervous system (CNS) myelinated nerve fibre is the remarkable product of the interaction between the axon and the myelin-forming cell, the oligodendrocyte. The axon is the long, slender projection of the neuronal cell body, along which electrical impulses travel to the synaptic terminus, while myelin is the lipid-rich insulating material that surrounds most CNS axons and increases the rate at which these impulses are transmitted. Together, the axon and its myelin sheath effect the timely relay of information between nerve cells, often over very long distances.

The formation of CNS myelin occurs through individual oligodendrocyte processes wrapping multiple times around a segment of axon, to construct a myelinated internode (Fig. 1). Oligodendrocyte processes are large, double-layered, membranous expansions. Cytoplasm is excluded from most of the area between the two layers of cell membrane, except at the periphery of the process where a cytoplasm filled channel remains. This arrangement of the membranous expanse of the oligodendrocyte process as it wraps around the axon gives rise to cytoplasm-free compact myelin, whereas non-compact myelin is formed where the channel of cytoplasm remains (Fig. 1). The cytoplasmic channel provides continuity between the myelin sheath and the oligodendrocyte soma (Ransom et al. 1991) and may also indirectly link the oligodendrocyte with the axon (Brahic & Roussarie, 2009).

Fig. 1.

CNS myelin is formed by oligodendrocyte processes wrapping multiple times around the axon. Schematic showing a single oligodendrocyte extending several processes, each forming a myelinated internode. Adjacent internodes are separated by small non-myelinated segments called nodes of Ranvier. The myelin sheath, which can be seen in the electron micrograph on the right, has two domains; compact and non-compact myelin. Ultrastructurally, the compact myelin consists of the oligodendrocyte cell membrane wrapped multiple times around the axon (Ax), forming the major dense line (between the intracellular surfaces of the process – represented in purple in the schematic) and the intraperiod line (between the extracellular surfaces of adjacent wraps). The inner (*) and outer tongues (**) of the oligodendrocyte process, together with the paranodal loops (not illustrated), constitute non-compact myelin. The most abundant proteins of CNS myelin, PLP/DM20 (green) and MBP (blue), are located in compact myelin. CNP (red) and sirtuin 2 (orange) are associated with non-compact myelin. MAG (purple) traverses the periaxonal space where it binds to receptors in the axolemma.

Although a range of axons of various diameters exists in all fibre tracts, in general, individual axons have a similar diameter along their entire length (Friede & Samorajski, 1970) (although small changes in size occur at the nodal and paranodal areas). The number of oligodendrocyte wraps (determining the thickness of the myelin sheath) and the width of the oligodendrocyte process (determining internodal length) are themselves positively dictated by the diameter of the axon in such a way as to maximise conduction velocity (Rushton, 1951). This precise matching of myelin sheath dimensions to axonal diameter reflects only one of several of the interdependent interactions between these two cells.

Along the length of an axon, adjacent myelinated internodes, which can extend for 1700 μm or more (Hess & Young, 1952; Murray & Blakemore, 1980; Hildebrand et al. 1993), are separated by small (∼ 1 μm), unmyelinated gaps, called nodes of Ranvier, which are rich in sodium channels and where action potentials are propagated. At the lateral edges of the internode, where the myelin sheath abuts the node, the cytoplasm-filled channel forms the regularly arrayed paranodal or lateral loops that form an attachment to the axolemma via transverse bands. This region constitutes the paranode. The juxtaparanode lies at the junction between the paranode and the compact myelin of the internode. Here the axolemma is rich in potassium channels. The precise structural and molecular composition of the nodal/paranodal region of the myelinated nerve fibre is crucial for saltatory impulse propagation and has been reviewed elsewhere (Brophy, 2003; Susuki & Rasband, 2008; Rosenbluth, 2009; Thaxton & Bhat, 2009).

The major proteins of CNS myelin

Myelin consists, by dry weight, of ∼ 70% lipid and ∼ 30% protein (Morell & Norton, 1980). The main lipids of myelin belong to three classes: cerebrosides (including their sulphate esters, sulphatides), other phospholipids and sterols (mainly cholesterol) (Wolman, 1968). In CNS myelin, proteolipid protein (PLP) and its smaller isoform DM20 constitute the most abundant protein component. Myelin basic protein (MBP) is the next most abundant protein. It was previously estimated, based on gel-derived proteomic analyses, that PLP/DM20 and MBP accounted for up to 45 and 35%, respectively, of all CNS myelin proteins. However, the recent identification of proteins hitherto not ascribed to myelin and mass spectrometry-based quantification studies suggest that their relative contributions to the total protein content are nearer 17 and 8%, respectively (Jahn et al. 2009). Mutations in the proteolipid protein (PLP1) gene encoding PLP/DM20 cause Pelizaeus–Merzbacher disease (PMD) and this review therefore focuses mainly on the role of PLP/DM20 in axonal-glial interaction, but other protein components of the myelin-enriched CNS tissue fraction will also be considered. These include myelin-associated glycoprotein (MAG), 2′,3′-cyclic nucleotide-3′-phosphodiesterase (CNP), sirtuin 2 (Fig. 1), myelin-associated oligodendrocytic basic protein (MOBP) and myelin oligodendrocyte glycoprotein (MOG).

PLP and DM20 are encoded by the highly conserved, X-linked PLP1 gene, which is expressed in the CNS predominantly in oligodendrocytes. These ∼ 30 and ∼ 25 kDa protein isoforms, whose primary structure is identical in man, mouse and rat (Macklin et al. 1987), are generated by alternate splicing of PLP1 pre-mRNA arising from differential usage of alternative splice sites in exon 3 (Milner et al. 1985; Kronquist et al. 1987; Nave et al. 1987; Simons et al. 1987). It is proposed that PLP, which is highly hydrophobic, constitutes a tetraspan transmembrane protein with interspersed extracellular and intracellular loops (Popot et al. 1991; Weimbs & Stoffel, 1992). DM20 differs from PLP in that it lacks 35 amino acids in the second (or intracellular) loop. PLP1 expression in the CNS commences prior to myelination, with the classical DM20 gene product dominating at this stage (Kronquist et al. 1987; LeVine et al. 1990; Schindler et al. 1990; Ikenaka et al. 1992; Timsit et al. 1992). As oligodendroglia contact axons and myelination ensues, PLP1 gene expression increases markedly and PLP becomes the predominant isoform (Kidd et al. 1990; Trapp et al. 1997). PLP/DM20 is enriched in compact myelin, where the two proteins can form a hetero-oligomeric structure (McLaughlin et al. 2002).

Novel Plp1 mRNAs and protein isoforms, expressed not only in oligodendrocytes but also in neurones, have been reported (Bongarzone et al. 1999). These soma-restricted isoforms (so named because of their localisation to the cell body) have so far only been demonstrated in the mouse. However, the results of a recent analysis of the human PLP1 gene coding sequence suggests the existence of neurone-specific PLP1 splice variants and protein isoforms in man (Sarret et al. 2010). Throughout this review, the terms PLP and DM20 are used to refer to the classic isoforms.

Pelizaeus–Merzbacher disease and spastic paraplegia type 2

In man, mutations in PLP1 cause PMD (Gencic et al. 1989; Hudson et al. 1989; Trofatter et al. 1989) and the allelic axonopathy, spastic paraplegia type 2 (SPG2) (Saugier-Veber et al. 1994). PMD is the prototypic leukodystrophy; representing a range of genetic disorders in which myelin formation and maintenance are perturbed. SPG2 belongs to a genetically heterogeneous group of neurological disorders, the hereditary spastic paraplegias (HSP), that are characterised by degeneration of distal parts of long spinal axons (DeLuca et al. 2004). Over 100 mutations in the PLP1 gene have been identified in man (Hudson et al. 2004) and these are associated with a broad spectrum of neurological disorders ranging from severe connatal forms of PMD (MIM 312080) at one end of the spectrum to pure forms of SPG2 (MIM 312920) at the other (reviewed in Saugier-Veber et al. 1994; Nave & Boespflug-Tanguy, 1996; Garbern & Hobson, 2002; Garbern, 2007). Mutations that have been identified include duplications/triplications of the gene (60–70%; intermediate to severe forms of PMD), deletions of the entire gene (1–2%; milder forms of PMD and SPG2) and point mutations, small deletions or insertions (15–20%; broad spectrum of clinical severity) (reviewed in Cailloux et al. 2000; Hudson et al. 2004; Garbern, 2007; Woodward, 2008). Within the last group, severe forms of PMD have been shown to associate with missense mutation in evolutionarily highly conserved regions of the protein, while milder forms of PMD are associated with substitutions of less conserved amino acids and with protein truncation (Cailloux et al. 2000). Notably, missense or nonsense mutations in the PLP-specific coding region, which would be predicted not to interfere with the sequence or expression of DM20, are associated with relatively mild forms of PMD or with SPG2 (Saugier-Veber et al. 1994; Osaka et al. 1995; Hodes et al. 1997).

The symptoms of PMD are related to widespread hypomyelination and include impaired motor development, nystagmus (abnormal eye movements), ataxia, choreoathetosis (abnormal movements of the upper limbs) and cognitive impairment. Life expectancy is reduced, and in severe forms, death may occur in infancy. In contrast, myelination may not be severely affected in SPG2 and symptoms may be limited to spasticity of the lower limbs (reviewed in Hudson et al. 2004). In some cases, the allelic disorders cannot be discriminated because complicated SPG2 and milder forms of PMD can share some clinical features, including nystagmus, cerebellar ataxia and pyramidal syndrome (Saugier-Veber et al. 1994). Incidentally, mutation in another myelination-associated gene, the fatty acid 2-hydroxylase (FA2H) gene, causes a form of complicated HSP (SPG35) with brain white matter abnormalities (Dick et al. 2010).

Spontaneously occurring mutations in the Plp1 gene in animals have provided a number of useful models of PMD/SPG2, which also span a broad clinical spectrum (reviewed in Yool et al. 2000; Nave & Griffiths, 2004). These include the shaking pup (Nadon et al. 1990), the myelin-deficient rat (Boison & Stoffel, 1989), jimpy (Nave et al. 1986) and rumpshaker mice (Schneider et al. 1992). Although there are no naturally occurring animal models of PMD caused by null mutation or gene duplication, mice lacking PLP/DM20 (Boison & Stoffel, 1994; Klugmann et al. 1997) and mice (Kagawa et al. 1994; Readhead et al. 1994) and rats (Bradl et al. 1999) harbouring multiple copies of the wild-type Plp1 gene have been generated using transgenic technology, providing useful models of SPG2 and PMD, respectively.

These animal models, combined with cell culture studies, have helped elucidate the molecular and cellular mechanisms underlying disease pathogenesis. Briefly, sequence changes predicted to cause protein misfolding and endoplasmic reticulum retention, lead to activation of the unfolded protein response (Southwood & Gow, 2001), alterations in the composition and structure of the myelin sheath and, in severe cases, apoptosis of large numbers of oligodendrocytes (reviewed recently in Campagnoni & Skoff, 2001; Nave & Griffiths, 2004; McLaughlin et al. 2006; Garbern, 2007; Woodward, 2008). Oligodendrocyte apoptosis also results when Plp1 is overexpressed, especially at high levels (Kagawa et al. 1994; Readhead et al. 1994; Anderson et al. 1999; Cerghet et al. 2001). In surviving oligodendrocytes, PLP/DM20 and cholesterol are sequestered to late endosomes/lysosomes and autophagosomes, perturbing myelin synthesis, composition and stability (Kagawa et al. 1994; Readhead et al. 1994; Simons et al. 2002; Karim et al. 2010). In contrast, the loss of function of PLP/DM20 in effective null mutants has relatively minor effects on myelin structure (see below), composition (Klugmann et al. 1997; Jurevics et al. 2003; Werner et al. 2007) or oligodendrocyte survival (Klugmann et al. 1997; Yool et al. 2001).

In this review we will focus on what we have learned about axon–glial interaction from animal models of PMD/SPG2, in particular from mice with functional null alleles or overexpressing the wild-type Plp1 gene. For wider reading the reader is referred to a number of recent reviews (Campagnoni & Skoff, 2001; Mobius et al. 2008; Woodward, 2008; Gow & Wrabetz, 2009; Fulton et al. 2010).

Roles for PLP/DM20 in myelination and in mediating oligodendroglial support of axonal function

Two lines of PLP/DM20-deficient mice have been generated using distinct gene targeting strategies (Boison & Stoffel, 1994; Klugmann et al. 1997). These mice have shown that myelin formation and maintenance does not require PLP/DM20, as in both models the myelin sheath still attains normal thickness. However, the sheath harbours various subtle abnormalities of the intraperiod line (Boison & Stoffel, 1994; Klugmann et al. 1997; Yool et al. 2002a; Rosenbluth et al. 2006), confirming earlier conclusions (based on the examination of animals harbouring point mutations in the Plp1 gene) that PLP/DM20 acts to stabilise the intraperiod line of compact myelin (Duncan et al. 1987, 1989). The intraperiod line represents the 20 Å (2 nm) space that lies between the extracellular surfaces of adjacent layers of the oligodendrocyte process as it wraps around the axon. Unlike the severely dysmyelinated Plp1 mutants (Arroyo et al. 2002), Plp1 knockout mice form apparently normal axon–glial junctions at the nodal region (Edgar et al. 2004b).

Although PLP/DM20-deficient myelin is abundant, a higher than normal proportion of small diameter axons fail to acquire, or are delayed in acquiring, a myelin sheath (Boison & Stoffel, 1994; Yool et al. 2001; Edgar et al. 2002). Interestingly, this abnormality is not observed in mice specifically lacking PLP (Spörkel et al. 2002), suggesting that DM20 is important in the early stages of axon recognition and/or wrapping of the oligodendrocyte process around small diameter axons. It is not known how DM20 might facilitate myelination of small diameter axons, but an indirect role in the modulation of oligodendroglial cytoskeletal dynamics is suggested by the fact that post-translational regulation of sirtuin 2 (SIRT2) is altered in Plp1 knockout mice (Werner et al. 2007). SIRT2, an NAD⁺-dependent protein deacetylase, regulates microtubule dynamics in developing oligodendrocytes (Li et al. 2007); a mechanism that may be particularly critical for bringing about the acute changes in shape that are required for the oligodendrocyte process to engage with the small diameter axon.

Despite the fact that PLP/DM20-deficient myelin harbours only subtle structural abnormalites, Plp1 gene knockout mice and PLP1 null patients develop a late-onset axonal degeneration, suggesting that Plp1 gene products mediate oligodendroglial support of axonal integrity (Griffiths et al. 1998; Garbern et al. 2002). That the genetically defective oligodendrocyte is sufficient to confer an axonopathy was confirmed using cell transplantation into shiverer mice (Edgar et al. 2004b).

Myelination in the mouse optic nerve begins around postnatal day 7 (P7). In Plp1 gene knockout mice, occasional amyloid precursor protein (APP) and phosphorylated c-Jun NH₂-terminal kinase (JNK) foci, markers of axonal transport stasis (Koo et al. 1990; Middlemas et al. 2003), can be detected by P30, using immunohistochemistry (Fig. 2). Electron microscopic examination shows that early axonal changes are associated with focal accumulation of axonal transport cargoes, mainly mitochondria, at the distal juxtaparanodal region of small diameter axons (Griffiths et al. 1998). Initially, organelles accumulate at the periphery of the axon (Fig. 1D in Edgar et al. 2004b) then gradually accrue, filling the entire diameter of the axon. Subsequently, long axons of the fasciculus gracilis (one of the longest fibre tracts in the mouse nervous system) degenerate in a manner akin to a dying back axonopathy, making this mouse a useful model of SPG2. While the mechanisms are not yet understood, impaired energy supply leading to defects in translocation of axonal transport cargoes has long been considered a potential mechanism of distal axonal degeneration (Spencer et al. 1979). P0 (the major structural protein of PNS myelin) or DM20 alone does not rescue the axonal phenotype (Stecca et al. 2000; Yin et al. 2006, 2008), highlighting the essential function of PLP in axonal support.

Fig. 2.

Early indications of axonal transport impairment in the optic nerve of the Plp1 gene knockout mouse. Immunohistochemical detection of amyloid precursor protein (APP) and phosphorylated c-Jun NH₂-terminal kinase (JNK) in longitudinal sections of the postnatal day 30 (P30) optic nerve of A and C wild type and B and D Plp1 knockout mouse optic nerve. Antibodies to APP (MAB348; Chemicon International, Temecula, CA, USA) and phosphorylated JNK (Cat. No. 9251; Cell Signalling Technology, Beverly, MA, USA) were visualised indirectly using chromagen diaminobenzidine and the paraffin sections were counterstained with haematoxylin to label cell nuclei. Myelination in the mouse optic nerve commences around P7 and the axonal changes depicted here, which arise secondary to the primary genetic alteration in the oligodendrocyte, are relatively rare in the P30 nerve. Scale bar: 30 μm.

Distal degeneration of axons of the fasciculus gracilis has also been observed in aged mice with both a hypo- (Edgar et al. 2004a) and a hypermyelinating phenotype (Harrington et al. 2010). However, the altered myelin volume associated with these mutations makes it difficult to evaluate the contribution of individual oligodendroglial components. In contrast, mice lacking CNP, an oligodendroglial protein residing in non-compacted myelin, produce an appropriately thick myelin sheath, but develop morphologically similar axonal changes to those reported in the Plp1 knockout mouse (Lappe-Siefke et al. 2003). However, axonal organelle accumulations occur earlier in the Cnp1 knockout mouse and are contemporaneous with morphological indications of axonal degeneration (Lappe-Siefke et al. 2003; Edgar et al. 2009). CNP-deficient myelin appears normal, except for swelling of the inner tongue process of some oligodendrocytes (Lappe-Siefke et al. 2003; Edgar et al. 2009) and disruption, in older animals, of the normal architecture of the paranodal region (Rasband et al. 2005).

Together, these observations demonstrate that the physical presence of myelin is, in itself, not sufficient to maintain axonal integrity. This conclusion is further supported by the fact that axonal degeneration, preceding or in the absence of inflammation or demyelination, has also been observed in mice deficient in MAG (Yin et al. 1998; Loers et al. 2004; Pan et al. 2005; Nguyen et al. 2009) or myelin 2-hydroxylated sphingolipids (Zoller et al. 2008). Conversely, axonal degeneration is not an inevitable consequence of ensheathment by a biochemically altered and/or structurally defective sheath, as indicated by its absence in mice lacking MOG, MOBP or MBP or with perturbation of the normal PLP/DM20 ratio (Griffiths et al. 1998; Yool et al. 2002b; Delarasse et al. 2003; Wang et al. 2008). Rather, it would appear that the oligodendrocyte plays a key role in supporting the integrity of the myelinated axon, a function that is distinct from its role in myelination (Nave & Trapp, 2008; Edgar & Nave, 2009; Nave, 2010). Some of the mechanisms by which oligodendrocytes could potentially support axonal function and integrity are summarised in Table 1.

Table 1.

Some of the known or putative mechanisms by which the oligodendrocyte could potentially mediate support of axonal integrity and/or function

Mechanism by which oligodendrocytes could mediate support of axonal function and integrity, under normal conditions or in response to injury	References
Sequestration of effete axonal organelles for degradation	Spencer & Thomas (1974)
Transfer of trophic factors from the oligodendroglial cytoplasm to the axon via double-walled, coated invaginations	Novotny (1984)
Stabilisation of the axonal cytoskeleton	Nguyen et al. (2009) and Yin et al. (1998)
Modulation of the activation status of axonal transport regulators	Morfini et al. (2009)
Perpetuation of metabolic support	Nave (2010) and Nave & Trapp (2008)
Shielding from inflammatory factors released by cells of the innate or adaptive immune system	Reviewed in Franklin & ffrench-Constant (2008) and Nave (2010)
Oligodendroglial peroxisomal function maintains white matter homeostasis	Kassmann et al. (2007)

What are the consequences for the axon if it is completely denuded (demyelinated)?

If the myelinated axon is dependent on the oligodendrocyte for support, and altered function of the oligodendrocyte has detrimental effects on the axon, what is the situation if the axon is completely denuded or demyelinated (i.e. the covering myelin sheath is lost completely)? Does this have implications for the demyelinated axon? This seemingly simple question is surprisingly difficult to answer because demyelination does not occur in isolation. Even when demyelination is chemically induced, cells of the immune system become ‘activated’ in response to the injury. In multiple sclerosis (MS), the best known demyelinating disorder of the CNS, active demyelination is inextricably coupled to inflammation, and the axonal changes that accompany demyelination cannot be dissociated from the inflammatory processes (reviewed in Lassmann, 2010).

Inflammation also accompanies demyelination in models of PMD. Unlike Plp1 knockout mice, which produce and maintain almost normal amounts of myelin, transgenic mice harbouring multiple copies of the wild-type Plp1 gene (models of PMD due to gene duplication) demonstrate premature arrest of myelination (dysmyelination), which is followed by spontaneous demyelination and neurodegeneration (Kagawa et al. 1994; Readhead et al. 1994; Inoue et al. 1996; Anderson et al. 1998, 1999). Demyelination in these models is probably triggered by a lack of stability of the biochemically abnormal myelin sheath (Simons et al. 2002; Karim et al. 2007, 2010). However, recent studies show that secondary immune responses also play a role (Ip et al. 2006, 2007, 2008; Kroner et al. 2010). The demyelinating CNS of hemizygous Plp1 transgenic mice (Plp1 tg; line #66; Readhead et al. 1994) contains markedly increased numbers of microglia/macrophages (Ip et al. 2006, 2008; Tatar et al. 2010) and small but significantly increased (cf. wild type), numbers of pathogenetically relevant CD8⁺ T cells (Ip et al. 2006) and, therefore, demyelination in these models does not occur in isolation either.

Axonal swelling and focal impairment of fast anterograde and retrograde axonal transport occur at sites of acute demyelination, coinciding with microglial/macrophage ‘activation’ in the optic nerve (Edgar et al. 2010) corpus callosum (Fig. 3) and other brain regions of another Plp1 gene overexpressing mouse (line #72; Readhead et al. 1994). In contrast, where demyelination is complete, axons generally appear intact. Although the cause and final outcome of the axonal changes in the inflamed regions is not yet known, there is evidence from other experimental systems to suggest that axonal injury could arise as a consequence of axono-toxic factors produced by invading lymphocytes and/or activated microglia/macrophages acting directly on the denuded axon (Koeberle & Ball, 1999; Medana et al. 2001; Garthwaite et al. 2005; Stagi et al. 2005; Takeuchi et al. 2005; Palin et al. 2008), or secondarily to microglial-mediated injury to the oligodendrocyte itself, before demyelination ensues (Howell et al. 2010). For example, it has been demonstrated in vitro that tumour necrosis factor alpha (TNF-α) and nitric oxide (NO), which are produced by ‘activated’ microglia, inhibit axonal transport via p38 stress-activated protein kinase or JNK-dependent phosphorylation of the anterograde molecular motor protein, kinesin (De Vos et al. 2000; Stagi et al. 2005, 2006).

Fig. 3.

Cyan fluorescent protein (CFP) expression in neurones reveals focal axonal swelling in the demyelinating CNS. Plp1 overexpressing mice (line #72) develop a slowly progressing demyelination. By crossing these mice with the Thy1-CFP line (Feng et al. 2000) (B6.Cg-Tg(Thy1-CFP)23Jrs/J) we were able to visualise axonal swelling in the demyelinating white matter. (A) A 10-μm-thick section through the corpus callosum of a postnatal day 90, Plp1 transgenic*Thy1-CFP mouse. CFP labelling was enhanced using immunohistochemical staining with a primary antibody to GFP (ab 6446; Abcam, Cambridge, UK), which was visualised with a goat anti-rabbit fluorescein-labelled secondary antibody. Focal swelling (arrows) of corpus callosal axons reach dimensions similar to those of the cell bodies in the cerebral cortex above (asterisk). (B) High power image of swellings in a different animal. (C) Microglia/macrophages labelled with an antibody to the pan-leukocyte marker CD45 (MCA1388; AbD Serotec, Oxford, UK) in the demyelinating corpus callosum appear ‘activated’. Scale bar (B,C): 20 μm.

Specialised requirements of the myelinated axon and pathways involved

Although many CNS axons are not myelinated (for example, 80% of axons in the basal telencephalic commissure in adult rhesus monkey are unmyelinated; LaMantia & Rakic, 1990) the majority of axons, which are over 0.2 μm in diameter, do acquire a myelin sheath (Hirano & Llena, 1995). While removal per se of previously formed myelin may not be detrimental to axonal survival, unless combined with other factors (reviewed by Dutta & Trapp, 2010), it has been proposed that myelination by itself could confer on the axon a particular requirement for oligodendroglial-mediated support (Nave & Trapp, 2008; Edgar et al. 2010; Nave, 2010).

A clue to the nature of this dependence possibly comes from the demonstration that the transportation of axonal cargoes is impaired in Plp1 knockout mice. Although the neuronal cell processes represented by axons are long in relation to processes of other cells within the body, the biosynthetic and degradative machinery still resides within the neuronal cell body. Materials, including mitochondria, therefore have to be actively transported by an ATP-dependent process that is driven by molecular motors, dynein and kinesin, along the axon in both directions and targeted to their (often distant) destinations. In the myelinated fibre there may be a requirement for the specific targeting of cargoes to nodal/paranodal and internodal domains that could add to the complexity of this process. Certainly, CNS myelin is known to influence the density (Mutsaers & Carroll, 1998; Andrews et al. 2006; Hogan et al. 2009) and motility (Kiryu-Seo et al. 2010) of axonal mitochondria, while oligodendrocytes influence the axolemmal localisation of ion channels (Kaplan et al. 1997; Baba et al. 1999; Rasband et al. 1999), which are probably also transported into the axon by molecular motors (Xu et al. 2010).

Retrograde transport studies in the Plp1 knockout model of SPG2 show that axonal cargoes are transported at an apparently appropriate rate, some distance along the axon, before some of these cargoes become ‘stuck’ at distal juxtaparanodal regions. In heterozygous Plp1^+/− mice, which are mosaics of wild-type and mutant myelin, axonal cargoes specifically accumulate at internodes myelinated by PLP/DM20-deficient oligodendrocytes, suggesting that the oligodendrocyte modulates axonal transport locally (Edgar et al. 2004b).

On the basis of this, it seems reasonable to hypothesise that the oligodendrocyte locally modulates the activity of motor protein subunits by regulating the activity of kinases (Morfini et al. 2009) in order to influence the delivery of transport cargoes to specific domains of the myelinated axon. Support for this putative function comes from the fact that oligodendrocytes locally modulate neurofilament phosphorylation at the nodal/paranodal junction of the myelinated axon (reviewed in Witt & Brady, 2000). Focal transport impairment in the Plp1 knockout mouse could potentially result from a localised dysregulation of motor protein subunit function or cargo binding (Hirokawa et al. 1990; Trinczek et al. 1999; Morfini et al. 2002; De Vos et al. 2003). For example, dynein activity is modulated through association with molecules such as Lis1 and Nudel. Nudel is itself a substrate for cyclin-dependent kinase 5 (Cdk5) and its activator p35. Deregulation of Cdk5 activity promotes neurodegeneration in vitro (Patrick et al. 1999) and the development, in mice, of organelle-filled axonal swellings (Bian et al. 2002) that appear morphologically similar to these observed in the Plp1 knockout mouse. That length-dependent distal axonal degeneration could result from a disruption to axonal transport is supported by the observation that SPG10 has been linked to a mutation in kinesin heavy chain gene, KIF5A (Reid et al. 2002).

Secondly, the myelin sheath, which encases almost the entire axon, could potentially act as a barrier to metabolic exchange between the extracellular space and the axon (Nave, 2010) by only allowing limited access of extracellular fluid to reach the axonal surface (Rosenbluth, 2009). This could place a requirement on the oligodendrocyte to monitor the local axonal energy status and convey axonal metabolic support (Nave, 2010), possibly utilising the NAD⁺-dependent deacetylase, sirtuin 2 (SIRT2) (Nave, 2010), a component of non-compact myelin and a potential sensor of axonal NAD⁺/NADH ratios which is absent from PLP/DM20-deficient myelin (Werner et al. 2007). The oligodendrocyte cytoplasmic channel coursing through the covering myelin sheath (and forming the non-compacted myelin), which is compromised in some CNP-deficient oligodendrocytes (Lappe-Siefke et al. 2003; Edgar et al. 2009), may have an important role as a route for perpetuating this putative support. The possibility that impaired transfer of materials between the oligodendrocyte soma and the axon–glial junction could deprive the axon of essential support is substantiated by a recent report showing progressive focal organelle accumulation and axonal degeneration in the taiep mutant rat (Wilkins et al. 2010), in which microtubule accumulation in oligodendrocytes has been shown to interfere with trafficking between the oligodendrocyte soma and the myelin sheath (Song et al. 2001).

Conclusion

The studies described above demonstrate how mouse models of PMD/SPG2 have contributed to what is currently known about the remarkable nature of the interdependent relationship between the axon and the oligodendrocyte. Current work focuses on determining the mechanism and molecules involved. The Plp1 mutants are also beginning to reveal how microglia and other cells of the immune system respond to the primary genetic defect in the oligodendrocyte and (potentially) impact axonal and oligodendroglial function and integrity.