Oligodendrocytes: biology and pathology

In the spinal cord, most oligodendrocytes derive from a specialized domain of the ventral ventricular zone, which first gives rise to motor neuron precursors, and then, after the neurogenic/gliogenic switch, to oligodendrocyte precursor cells/progenitors (OPCs) [107, 181, 183, 214]. From there, OPCs migrate all through the spinal cord and finally differentiate into myelin-forming oligodendrocytes. Later, an additional source of OPCs arises in the dorsal spinal cord, contributing to 10–15% of the final oligodendrocyte population in the spinal cord [25, 49, 195].

In the forebrain, the first OPCs originate in the medial ganglionic eminence and anterior entopeduncular area of the ventral forebrain. These OPCs populate the entire embryonic telencephalon including the cerebral cortex, and are then joined by a second wave of OPCs derived from the lateral and/or caudal ganglionic eminences. The third wave of OPCs, finally, arises within the postnatal cortex [86]. These different populations of OPCs are functionally redundant: When any one of them is destroyed at source by the targeted expression of a toxin gene in mice, the remaining cells spread into the vacant territory and restore the normal distribution of OPCs. As a result of this, a normal complement of oligodendrocytes and myelin can be produced, and the mice develop, survive, and behave normally [86]. In spite, or maybe even due to this functional redundancy, the different OPC populations seem to be fierce competitors.

In the developing rodent spinal cord, the competition is won by OPCs derived from the ventral ventricular zone: These cells represent the first wave of cells of the oligodendrocyte lineage generated in the spinal cord, and eventually give rise to 85–90% of the final oligodendrocyte population found in this organ [25, 49, 195]. In this case, competition for limiting quantities of growth factors like platelet-derived growth factor (PDGF) might be a determining factor for winning the competition [9, 26, 196].

In the developing forebrain, however, the situation seems to be quite different. Here, the competition is clearly lost by the first wave OPCs: although these cells are the first to distribute and to occupy vacant territories, their contribution to the total OPC population rapidly declines, until they are almost completely eliminated from the adult forebrain.

The reasons for the eventual loss of one OPC population in the developing forebrain remain completely unclear, and one can only speculate about possible causes and consequences [157].

According to such speculations, the first wave of forebrain OPCs emerging from the medial ganglionic eminence and the anterior entopeduncular area might represent the most “primitive source” of oligodendrocytes, a relic that lost its importance during evolution, when new sources of OPCs developed in the expanding brain. This assumption is supported by the observation that birds, which have much less cortical volume than mammals derive their OPCs only from the anterior entopeduncular area [157]. Alternatively, the first wave OPCs could be lost in the course of a (not yet demonstrated) oligodendrocyte turnover, or as a result of changes in the availability/responsiveness to growth and differentiation factors. Moreover, they might become dispensable if they have functions during development, that are not required in the adult [157].

No matter whether OPCs belong to the first, second, or third wave of cells, they have one point in common: They have to travel long distances in order to end up in their final place of destination. This migration is tightly controlled.

To date, three different classes of secreted molecules seem to be involved in the migration of OPCs: growth factors like PDGF, FGF [155, 174] or hepatocyte growth factor [211, 138]; chemotropic molecules like netrins and secreted semaphorins [74, 193]; and the chemokine CXCL1 [37]. Although there is no doubt that these factors play a role in OPC migration, the exact mode of action of these factors is still a matter of controversy, in part due to differences in experimental models, culture systems, and time points studied.

OPC migration is not only controlled by secreted molecules, but is also regulated by contact-mediated mechanisms involving many different extracellular matrix proteins and cell surface molecules [52, 88, 199], N-cadherins [125, 133, 145, 167, 190], and possibly even additional, yet unidentified molecules.

From all these different molecules, a common theme evolves, demonstrating contact-based migration of OPCs over extracellular matrices, axonal tracts, and astrocytic surfaces [37]. Once located at their final destination, some OPCs persist into adulthood (see Appendix 1), while the vast majority differentiates to myelin-producing oligodendrocytes.

This process involves signaling processes between the Notch1 receptor, its ligand Jagged 1 located on the axonal surface [55], and γ-secretase [150, 202]. Interestingly, oligodendrocytes have only a brief period of time for myelination early during differentiation, and are relatively incapable of myelinating once they are mature [202]. Moreover, the ensheathment of multiple axons by a single oligodendrocyte is a highly coordinated event: oligodendrocytes do not ensheath different axons sequentially at different time points, but are done within a brief window of time, typically within 12–18 h [8, 202].

Oligodendrocytes select axons with diameters above 0.2 μm [171]. The molecular cues for this recognition remain unknown. In the peripheral nervous system, the critical axonal signal for myelination by Schwann cells is provided by neuronal neuregulin-1 (NRG1) type III, interacting with glial ErbB receptors, and it was thought for a long time that NRG1/ErbB signaling might also regulate myelination in the CNS. This point of view was supported by the observation that not only Schwann cells, but also oligodendrocytes express erbB receptors (oligodendrocytes erbB2 and erbB4, Schwann cells erbB2 and erbB3, but only little erbB4 [197]). It was further observed, that oligodendrocytes fail to develop in spinal cord explants derived from mice lacking neuregulin [198], that erbB2 is required for the development of terminally differentiated spinal cord oligodendrocytes [141], and that erbB2 plays a role in governing the properly timed exit from the cell cycle during development into myelinating oligodendrocytes [89]. Moreover, mice with dominant negative erbB4 have more differentiated oligodendrocytes, but each of them seems to myelinate less axonal surface and to produce thinner myelin sheaths than its wild-type counterpart [159]. And last, mice haploinsufficient for type III NRG1 are hypomyelinated in the brain, and normally myelinated in the optic nerve and spinal cord [187]. However, recent observations challenge the assumption that NRG1/ErbB signaling might also regulate myelination in the CNS. Instead, they suggest that NRG1/ErbB signaling serves distinct functions in myelination of the peripheral and central nervous system [19]: Mice completely lacking NRG1 beginning at different stages of neural development assemble normal amounts of myelin, on schedule [19]. When neuregulin signaling is completely abolished in oligodendrocytes of double mutants lacking ErbB3 and ErbB4 (and carrying ErbB2 without ligand-binding activity), CNS axons are nevertheless myelinated without delay, and at the same level as in control mice, at least until postnatal day p11, the latest time point studied [19]. However, while myelination was not delayed even in the absence of NRG1, there was clear evidence for premature, and even for hyper-myelination in NRG1 type III overexpressing mice [19]. Cumulatively, these data suggest that normal myelination occurs independently of NRG1 signaling in vivo. They also demonstrate that excess NRG1 can initiate the myelination program in CNS development, but that this function is normally provided by different, still unknown axonal signaling systems.

The data summarized above led to the conclusion that the onset of CNS myelination in normal development might be determined by the degree of neuronal differentiation, and not by the timing of an intrinsic oligodendrocyte differentiation program [19].

One essential signal for the onset of myelination seems to be provided by the electrical activity of neurons. For example, the optic nerves of mice which had been reared in the dark developed fewer myelinated axons than control optic nerves [62], optic nerves of naturally blind cape-mole rats are hypomyelinated [139], and blockade of sodium-dependent action potentials in developing optic nerves inhibits myelination [38]. Vice versa, increasing neuronal firing with α-scorpion toxin enhances myelination [38], and premature eye opening accelerates myelination in rabbit optic nerves [186]. Action potential firing leads to the release of ATP [177] and adenosine [96, 115], which can mediate neuron-glial communications. In the CNS, adenosine inhibits the proliferation of OPCs, stimulates their differentiation, and promotes the formation of myelin [178]. The ATP released from axons firing axon potentials does not directly act on OPCs or oligodendrocytes. Instead, it triggers the release of leukemia inhibitory factor (LIF) from astrocytes, which promotes myelination by mature oligodendrocytes [67]. This fact could explain why LIF−/− mice display impaired myelin formation [23], and why interfering with astrocyte biology leads to myelin abnormalities, as seen in mice with knock-out of the astrocytic protein GFAP [104], and in patients with Alexander disease, a fatal white matter disorder of childhood caused by mutations in the GFAP gene [124].

Although the electrical activity of neurons in the CNS is an essential promyelinating factor, additional changes on axons seem to be needed to drive efficient myelin formation. Some of these changes seem to be induced by electrical activity as well. For example, during development, all growing nerve fibers express the polysialylated-neural cell adhesion molecule (PSA-NCAM), which prevents homophilic NCAM–NCAM adhesion and, more generally, cell–cell interactions. Only when this form of NCAM is downregulated, as it is in electrically active neurons [91, 97], myelination can proceed [32]. Another molecule correlating with the extent of oligodendrocyte maturation and myelination is LINGO-1, a transmembrane protein with leucine-rich repeats and an immunoglobulin domain expressed in neurons and oligodendrocytes. Loss of LINGO-1 function in oligodendrocytes leads to increased myelination, whereas its overexpression inhibits myelin formation [122].

There are certainly more, yet unidentified molecular mechanisms enabling oligodendrocytes to recognize, ensheath, and wrap axons [8].

The window of time available for the onset of myelination seems to be very narrow, in the range of 12–18 h during which oligodendrocytes have to wrap layer after layer of plasma membrane around multiple axons [8]. Hence, they have to synthesize, sort, and traffick an enormous amount of proteins in short time. The machinery for these processes is rather complex. For example, one major myelin protein, myelin basic protein (MBP, Table 1), is targeted by transport of its mRNA. The MBP mRNA is assembled into granules in the perikaryon of oligodendrocytes, transported along processes, and then localized to the myelin membrane [2, 3]. The other major protein, proteolipid protein (PLP)/DM20 is transported to myelin by vesicular transport through the biosynthetic pathway. Two different observations suggest that both delivery systems might be under the control of neuronal signals.

First, oligodendrocyte cultures without neurons express MBP, PLP, and galactosylceramide (GalC). However, under these conditions, PLP is predominantly intracellularly localized, and shows only little co-localization with MBP or GalC in the membrane sheets. In coculture with neurons, however, PLP, MBP, and GalC are co-localized. This rearrangement of the oligodendroglial plasma membrane is critically dependent on the presence of MBP. MBP seems to act as a lipid coupler by bringing the different layers of myelin in close position (a process termed “vertical membrane coupling”), and by clustering the lipid bilayer in lateral dimensions (“horizontal membrane coupling”). This neuron- and MBP-dependent clustering mechanism might be responsible for the concentration of membrane components within myelin [48].

Second, in the absence of neurons, PLP is produced by oligodendrocytes, but immediately internalized by endocytosis. After receiving neuronal signals, the rate of endocytosis is reduced, and PLP is transported from the late endosomes/lysosomes to the plasma membrane by exocytosis [192]. This regulation of PLP trafficking might represent a mechanism to store membrane produced in advance, before the onset of myelination, and to release this membrane on demand in a regulated fashion [171] (Fig. 1).

If the inciting injury is simultaneously directed against myelin and oligodendrocytes, sharply demarcated plaques of primary demyelination are induced [180]. Myelin sheaths are completely lost, while oligodendrocyte cell bodies may be partly preserved within active lesion areas. The paradigmatic example for such a mechanism is demyelination triggered by specific antibodies against MOG, an antigen expressed in highest density at the peripheral processes of oligodendrocytes covering the myelin sheath [22]. Upon binding of these antibodies, myelin sheaths are disintegrated by vesicular dissolution in case of massive complement deposition [54] or by phagocytosis of myelin fragments by macrophages (antibody dependent cellular cytotoxicity [21]. Acute injury of oligodendrocytes follow the pathway of necrosis. However, mature oligodendrocytes which have lost their myelin sheaths but survived the initial attack are then slowly removed from the lesions by apoptosis [208]. Following these initial changes, sharply demarcated focal demyelinated lesions are formed. A similar pattern of lesion formation is also seen when myelin and oligodendrocytes are destroyed through antigen-specific cytotoxic T-cells [164] or non-selectively by toxic products of activated macrophages (bystander demyelination). Under the latter conditions, however, the demyelinated lesions are associated with much more widespread damage to other cellular components as well, in particular axons.

Distinct types of lesions are seen, when the primary injury is due to a metabolic disturbance of oligodendrocytes, not directly affecting myelin. In this situation, demyelination is frequently incomplete. Such lesions not only contain areas of complete demyelination, but also diffuse myelin pallor is observed. At the edges of the lesions, a moth eaten pattern of demyelination is observed which may reflect the loss of single oligodendrocytes with their respective myelin sheaths. This pattern of demyelination is mainly seen in conditions of infections, which target oligodendrocytes, such as progressive multifocal leukoencephalopathy. In such conditions, virus antigen is abundant in oligodendrocytes and the cells are destroyed both by apoptosis or necrosis [129]. Toxic damage of oligodendrocytes, for example by cuprizone results in similar patterns of demyelination [191].

Energy deficiency in the white matter leads to a fundamentally different pattern of tissue injury. Also in this condition, oligodendrocytes are highly vulnerable, but they die by a process termed “distal/dying back oligodendrogliopathy”. In initial lesions, the most severely damaged parts of the cells are the most distal (periaxonal) oligodendrocyte processes [1]. This is reflected by a selective loss of proteins, which are predominantly located in this location, such as myelin-associated glycoprotein (MAG) and cyclic nucleotide phosphodiester [1, 68]. Conventional staining for myelin with luxol fast blue shows a diffuse or focal myelin pallor, while immunocytochemistry for proteins located within compact myelin (myelin basic protein or proteolipid protein) is unaffected. With progression of the lesions the majority of oligodendrocytes reveal nuclear condensation and in part nuclear fragmentation in the absence of the expression of activated caspase 3 (caspase independent apoptotic like cell death). Additional characteristic features of such lesions are the preferential destruction and loss of small caliber axons and a remarkable preservation of axons and myelin around larger blood vessels (arterioles and veins). Such tissue changes are the hallmark of initial ischemic lesions of the white matter, and occur within the first hours or days in a white matter stroke lesion [1]. Similar lesions, however, can also be seen in severe inflammatory brain lesions, for example in a subset of patients with acute multiple sclerosis or with virus infections of the white matter like herpes simplex virus encephalitis, cytomegalovirus encephalitis, or progressive multifocal leukoencephalopathy [1]. In the latter lesions, energy deficiency is associated with profound mitochondrial damage, which may at least in part be induced by disturbance of the mitochondrial respiratory chain through reactive oxygen and nitrogen species [114]. Upregulation of molecules which are induced by (hypoxic) tissue preconditioning, such as hypoxia inducible factor 1 alpha or stress proteins in the periphery of such lesions may exert local neuroprotective effects. Moreover, tissue areas with increased resistance to energy deficiency may alternate in the periphery of the lesions with more vulnerable areas, and can give rise to concentric layering of demyelinated and preserved tissue zones typically found in the lesions of Balo’s type of concentric sclerosis [175] (Fig. 2).

It has to be emphasized that these distinct patterns of demyelination segregate well in the initial stages of lesion formation and in patients with rapidly progressive white matter disease. However, in more slowly expanding lesions, these morphological features may in part be lost. Then, it may become difficult to determine the mechanism of tissue injury purely on morphological grounds. The final outcome of all the lesions described above is focal or diffuse areas of primary demyelination in the white matter.