Myelin Formation and Plasticity in the Central Nervous System

1. Introduction

Myelin is a multilamellar stack of membranes that envelops specific segments of neuronal axons within the nervous system [1]. Myelin in the central nervous system (CNS) is produced by oligodendrocytes (OLs). In contrast, the peripheral nervous system (PNS) relies on Schwann cells for this function [1, 2]. This membrane structure is among the most remarkable in the body; it is distinguished by a unique composition of lipids and proteins, setting it apart from typical cellular membranes [2–4]. In contrast to conventional cell membranes, it comprises of a higher proportion of lipids (≈75%) and a lower proportion of proteins (≈25%) [1–3]. Galactosylceramides and sulfatides, comprising of long-chain fatty acid groups, are the most typical myelin lipids. The high proportion of saturated, long-chain fatty acids exerts a significant impact on membrane structure, thereby influencing parameters such as myelin thickness and the packing density of lipids within myelin [2].

The discovery and understanding of myelin evolved over several centuries [5]. The first documented instance of the term “white matter” is attributed to Vesalius in the sixteenth century, while the observation of myelinated fibers is believed to have been made by van Leeuwenhoek in 1717. Notable contributions to this field include Remak’s axon theory (1836), Schwann’s description of nuclei (1839), and Virchow’s coinage of the term “myelin” in 1854. In 1872, Ranvier proposed the hypothesis that myelin originated from fatty cells, which would subsequently be recognized as Schwann cells [5]. The term “oligodendrocytes” was coined by del Río-Hortega in 1919 and formally defined in 1921 [5]. Subsequent research, notably that of Penfield in 1924, established these cells as equivalent to Schwann cells within the CNS. The twentieth century saw significant progress in the understanding of myelin, with its fatty composition, crystal-like structure, and functional role in saltatory conduction being clarified in 1954 [5]. Subsequent electron microscopic analysis provided the initial evidence supporting the hypothesis that Schwann cells envelop axons, thereby forming compact myelin spirals. The central origin of myelin was subsequently elucidated, as OLs are distinctly located away from axons; in 1962, Bunge provided definitive evidence that OLs are responsible for myelin production [5].

In view of the advances made in the field of cellular and molecular biology, subsequent research has focused on the molecular mechanisms of OL development and myelination, which have consequently become central topics in glial research [2]. A wide range of models, including human, rodent, zebrafish, Xenopus, and non-human primates, has been employed to study myelination. For a considerable period, OLs and other glial cells were regarded as a basic “support system.” However, recent advancements have prompted a reevaluation of these cells, leading to their recognition as active and essential participants within the CNS. The complex interactions among OLs, other glial cells, and neurons are now recognized as being critical for optimal brain function [2, 4, 6].

In the CNS, myelin and OLs are affected in several conditions (Figure 1) [7]. Myelin is a particularly susceptible target of CNS injuries, including hypoxia and ischemic stroke [2, 7]. However, the most prevalent pathology associated with OLs and myelin is MS, an inflammatory disease that predominantly affects young individuals worldwide [2]. It manifests with a higher frequency in women than in men, and the etiology of the disease is only partially understood. The current hypothesis is that genetic and environmental factors contribute to the development of MS [1, 2]. The immune system attacks OLs and myelin, and the CNS often fails to repair this damage, ultimately resulting in neuronal degeneration and severe disabilities [2]. Another group of disorders associated with myelin is referred to as leukodystrophies. While they have generally been regarded as genetic anomalies resulting in hypomyelination or demyelination, only a limited number of the implicated genes are exclusively expressed in OLs [1, 2].

Recent evidence highlights the active involvement of OLs and myelin in Alzheimer’s disease (AD). Despite the established knowledge that myelin is subject to alterations during the process of aging, the prevailing paradigm regarding OLs has historically portrayed them as passive entities within the context of AD pathogenesis [8]. However, new findings indicate that there is a functional shift in the orientation of OLs towards a disease-associated state in response to amyloid and tau lesions [8]. Rather than being inert, they play a dual role in AD pathogenesis, on the one hand providing potential protective responses against pathology, while on the other hand contributing to disease progression [8].

Myelin has also been implicated in neurodevelopmental and psychiatric disorders, including autism spectrum disorder (ASD) and schizophrenia [9]. It is noteworthy that rodent models displaying ASD mutations exclusively in oligodendroglia exhibit certain phenotypes that mirror those observed in individuals with ASD [10].

Here, we present a comprehensive overview of the mechanisms underlying myelination in the CNS and the key roles played by oligodendroglial cells across different contexts. While the primary focus of this review is on the CNS, occasional references to the PNS are included when structural or functional aspects of myelin are considered. Additionally, the emerging concept of myelin plasticity in the CNS is discussed, as it has become a major focus of research in the field [4, 6].

2. The journey from OPCs to myelin

OLs are derived from oligodendrocyte precursor cells (OPCs), which themselves originate from neural stem cells (NSCs) [1]. During the developmental process, OPCs undergo a series of changes that result in their differentiation into pre-myelinating OLs (Figure 2). These cells subsequently reach full maturity and functionality as fully mature OLs [1, 2], which extend processes that contact axons and generate the myelin sheath (Figure 2).

In the rodent cortex, OPCs emerge from the neuroepithelial regions surrounding the ventricles and are generated in three distinct developmental waves (E12.5, E15.5, and P0). It has been observed that these waves originate in the ventral region and progressively shift toward the dorsal regions [11, 12]. The OPCs from the first wave are predominantly replaced by medial and dorsal OPCs by postnatal day 10 (P10) [11]. In the spinal cord, two waves are identified: a ventral wave commencing at E12.5, which generates 80–90% of the total OPC population, and a dorsal wave that contributes the remaining 10–20% [13].

Once generated, OPCs migrate toward their target regions, guided by gradients of signaling molecules, extracellular matrix proteins (e.g., laminin, fibronectin), axon guidance cues (e.g., netrin-1, CXCL1), and growth factors such as vascular endothelial growth factor (VEGF) and platelet-derived growth factor (PDGF) [13]. Moreover, it has been shown that neuronal activity can promote OPC migration [14]. OPCs have been shown to express AMPA and NMDA receptors and to receive direct synaptic input from neurons [14, 15]. This finding lends support to the hypothesis that neurotransmitter release may function as a positional cue during migration. Furthermore, OPCs frequently migrate along blood vessels, using these vessels as structural “tracks” that facilitate the migration of OPCs between regions [16].

Upon reaching their destination, OPCs proliferate and differentiate [13]. The progression of these cells into pre-myelinating OLs and then into mature OLs is influenced by the activity of differentiation-promoting transcription factors, including myelin regulatory factor (Myrf). Mature OLs have been shown to initiate myelin production by expressing key myelin proteins, including MOG, MBP, and PLP. Together, these proteins facilitate the assembly of myelin sheaths, forming a compact bilayer lipid membrane that insulates and protects axons [13]. Myelin compaction is primarily mediated by MBP, which acts as a “zippering” protein between adjacent bilayer membranes [2, 17]. The process of MBP mRNA transport along microtubules from the OL soma to its processes, followed by on-site protein synthesis, is referred to as local translation [18]. It has been demonstrated that MBP exhibits a high degree of affinity for phosphatidylinositol-4,5-bisphosphate (PIP₂), a lipid constituent of the membrane [19]. The binding and neutralization of two PIP2 molecules by MBP result in the convergence of opposing bilayers, thereby forming the major dense line (Figure 2) and extruding cytoplasm from the space between them [20, 21]. PLP, a transmembrane protein, contributes to the adhesion of the extracellular leaflets of the myelin membrane (double intraperiod line, Figure 2), thereby further stabilizing the compact structure [21]. Mice deficient in PLP can still form compact myelin; however, the sheaths are physically unstable [22]. Despite its compaction, myelin facilitates the exchange of metabolites through cytoplasmic channels that are maintained by 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNP). CNP counteracts MBP-mediated compaction by binding actin and creating intracellular “pillars” that maintain open, narrow cytoplasmic spaces [23]. These channels are more abundant in young animals than in adults [24] and enable the transport of metabolites, vesicles, and trophic molecules to the periaxonal space, supporting axo-glial unit maintenance [2]. Other myelin proteins, such as myelin oligodendrocyte glycoprotein (MOG) and myelin-associated glycoprotein (MAG), have also been shown to contribute to the stability and integrity of the sheath [25, 26].

Myelin formation exemplifies a particularly dramatic instance of morphological change in a living cell [27, 28]. OLs are subject to a process of extensive cytoskeletal remodeling, which is necessary for the extension of their processes, the wrapping of their axons, and the production of myelin. During OL development, the process of actin polymerization has been shown to drive outgrowth, branching, and ensheathment [27, 28]. Conversely, the process of wrapping during myelination requires the disassembly of actin. Despite the broad consensus on the necessity of actin depolymerization, the precise dynamics of this process remain a subject of debate. A body of research posits a continuous cycle of turnover of polymerization and depolymerization as the underlying mechanism that generates the forces necessary for the process of wrapping [27]. In contrast, other studies propose that depolymerization occurs only once at the onset of wrapping [28]. A recent study found that the transition from actin polymerization to depolymerization requires inhibition of the P21-Activated Kinase 1 (PAK1), mediated by the binding of the NF2/Merlin protein [29]. These findings reveal the fascinating antagonistic interplay between the pro-oncogenic kinase PAK1 and the tumor suppressor NF2/Merlin in myelin membrane expansion.

Microtubules (MTs), the other major component of the cytoskeleton in OLs, also play critical roles in myelination. These structures can be categorized into two distinct types: radial MTs, which extend toward processes, and lamellar MTs, which spiral longitudinally around the myelin sheath and traverse cytoplasmic channels [24]. The function of MTs is critical for the transport of mRNA and vesicles from the soma to the processes and within the myelin sheath (as noted above). Additionally, MTs are involved in process extension via polymerization and depolymerization dynamics [30]. Recent research has demonstrated that the nucleation of MTs outside the cell body is essential for the process of myelination. In particular, tubulin polymerization-promoting protein (TPPP) has been shown to mediate MT nucleation at Golgi outposts within OL processes [31]. The in vivo studies on TPPP depletion havent revealed that TPPP-deficient mice exhibit shorter and thinner myelin sheaths, along with motor coordination deficits on the rotarod test [31].

The CNS is a highly complex network comprising a diverse array of neuronal types with varying morphologies and sizes across different brain regions. Even within a single region, an OL may encounter multiple neuron types that vary in axonal diameter and function. Yet, the OL exhibits the capacity to selectively myelinate certain axons while leaving others unmyelinated [3]. OLs possess an intrinsic capacity to form myelin, with the ability to generate up to 80 sheaths on different axons [2]. In the PNS, axons with diameters greater than 1 µm are typically myelinated [32]. In contrast, the process of CNS myelination does not appear to adhere to such a rigid size criterion. For instance, in specific brain regions, axons measuring as small as 200–300 nm in diameter are myelinated, while in other regions, these axons remain unmyelinated [3]. Specific biophysical cues have been shown to influence the permissiveness of axons to myelination. In zebrafish, the first axon to be myelinated is the large-caliber Mauthner axon, and the generation of additional Mauthner axons alters how individual OLs allocate their myelinating capacity [33]. In vitro, axonal geometry has been demonstrated to influence OL behavior. Specifically, high-curvature surfaces have been demonstrated to be non-permissive, while large-diameter, low-curvature axons have been observed to promote myelination [34]. Increasing the diameter of normally unmyelinated cerebellar granule cell axons in vitro has been demonstrated to be sufficient to trigger myelination [35]. These findings suggest that, as in the PNS, large axons in the CNS may present cues that attract OLs. However, it is important to emphasize that axonal size, in isolation, does not serve as a definitive indicator of a specific condition. Molecular cues have been demonstrated to play a role in this process. In the PNS, neuregulin 1 (Nrg1) has been demonstrated to be sufficient to attract Schwann cells and initiate myelination; however, in the CNS, Nrg1 disruption does not block myelination, suggesting that multiple cues are involved [13, 34].

Furthermore, emerging evidence has demonstrated the critical function of neuronal activity in directing OL targeting. Indeed, studies of the zebrafish spinal cord have revealed that the inhibition of action potentials (AP) or activity-dependent vesicle release results in a reduction of myelination in specific axons [36, 37]. The release of glutamate has been identified as a mediator of this effect in vitro [38], although the precise receptor subtypes involved remain to be elucidated. The occurrence of activity-dependent regulation is not uniform; some neurons control myelin targeting in one brain region but not in another [39]. It is noteworthy that neurons are not completely myelinated, as OLs generally avoid myelinating dendrites and the soma [40]. Furthermore, in the adult CNS, a significant number of axons persist in an unmyelinated state or exhibit discontinuous or incomplete myelination, even in the presence of OLs [41]. These observations suggest the potential involvement of repulsive signals. Eph–ephrin interactions have been demonstrated to induce OL process retraction and reduce myelination [42, 43]. Polysialylated neural cell adhesion molecule (PSA-NCAM) has also been demonstrated to exert repulsive effects, with increased myelination resulting from its blocking and decreased myelination resulting from its overexpression [44, 45]. Recent findings indicate that junctional adhesion molecule 2 (Jam2) functions as an inhibitory cue, restricting myelination of the somatodendritic compartment [46]. In summary, OL axonal selection is likely governed by an integrated code of biophysical properties, attractive molecular signals, repulsive cues, and both activity-dependent and activity-independent mechanisms. The relative contribution of each factor is likely to vary according to brain region, neuron type, and even local microenvironment, and the same cue may exert different effects in different contexts [13].

3. Functions of oligodendroglial cells and myelin

The primary function of myelin is to increase the speed of AP propagation along axons. Myelinated segments (internodes, Figure 2) are interrupted by short unmyelinated regions known as the nodes of Ranvier, which are flanked by paranodes and juxtaparanodes [2] (Figure 3). The nodes, defined as the gaps between two internodes, have been observed to contain a high density of voltage-gated sodium channels, predominantly the Nav1.6 isoform, within the adult mammalian nervous system [47]. Furthermore, they have been found to harbor other ion channels, including Kv7.2/Kv7.3 voltage-gated potassium channels, which have been demonstrated to contribute to the stabilization of the resting membrane potential, as well as Kv3.1b channels, which have been shown to facilitate high-frequency firing [47]. In juxtaparanodes, there is a high density of delayed rectifier voltage-gated potassium channels (Kv1.1, Kv1.2) along with their auxiliary subunit Kvβ2. It is hypothesized that these channels and subunits contribute to maintaining the internodal resting potential [47]. The paranodes, which are devoid of ion channel clusters, are nevertheless essential for axon–glia junctions. At this site, a tight adhesion is formed between the myelin sheath and the axolemma. Neurofascin-155 (Nfasc155) on the myelin membrane interacts with Caspr 1 and contactin on the axolemma [48]. Consequently, nodes of Ranvier have been shown to facilitate rapid propagation of APs and stabilize axon-glia units by segregating ion channels into non-myelinated spots and establishing thight interaction with axolemma. In saltatory conduction, the AP effectively "jumps" from one node of Ranvier to the next, while the high compaction of the myelin sheath prevents ionic leak that could disrupt signal propagation, thereby optimizing CNS conductivity (Figure 3).

The other essential function of OLs and myelin is their role in providing metabolic support to neurons (Figure 3). Experimental studies have identified a lactate transporter in the innermost layer of myelin, positioned in close proximity to the axon [2, 49, 50]. Blocking this pathway leads to hypomyelination and axonal degeneration, underscoring the essential role of lactate transfer from OLs in sustaining myelinated neurons. This metabolic coupling necessitates the sensing of axonal activity by OLs [49]. Accordingly, glutamate receptors have been identified in the innermost layer of brain myelin [49]. Activation of these receptors has been shown to promote GLUT1 trafficking to the outer myelin and paranodal regions, thereby increasing glucose uptake and enhancing lactate delivery to active axons [49].

While oligodendroglial cells (OPCs and OLs) are predominantly recognized for their role in myelination, numerous studies over the past years have highlighted their involvement in immune functions [51] (Figure 3). Within the CNS, the immune response is primarily orchestrated by microglia, the resident immune cells of the brain [2]. These cells perpetually survey their environment, poised to respond to injury or the intrusion of foreign substances.

OPCs have the capacity for phagocytosis, which serves to clear away debris, and they can also release cytokines, orchestrating responses in other cells [52]. OPCs exhibit several features that are analogous to those of microglia, including motility, phagocytic activity, and the capacity to respond to and release cytokines [52]. In vitro, primary rat OPCs have been demonstrated to phagocytose a variety of cellular debris, including myelin fragments and even cell membranes derived from astrocyte cultures [53]. In vivo evidence has emerged demonstrating the presence of axonal fragments and phagosomes within OPCs. Single-nucleus RNA sequencing data have revealed the expression of key phagocytic genes in OPCs, suggesting a potential role in brain shaping during development [54]. Beyond the clearance of debris, OPCs have also been demonstrated to contribute to the maintenance of CNS homeostasis under physiological conditions [51]. This process is facilitated by the release of TGF-β2, which has been shown to upregulate the microglial receptor CX3CR1 and promote a neuroprotective phenotype in microglia [55]. It has been demonstrated that depletion of NG2 glial cells consistently reduces the microglial homeostatic signature, leading to heightened activation and increased production of pro-inflammatory factors such as interleukin (IL)-1β, IL-6, and tumor necrosis factor (TNF) [56].

While oligodendroglial cells have not heretofore been regarded as a component of the immune system, OPCs have been demonstrated to exhibit a response to inflammation. In both human MS tissue [57] and mouse models [58], OPCs have been observed to express genes involved in antigen processing and presentation. These cells have been shown to produce a variety of inflammatory molecules, including cytokines and major histocompatibility complex (MHC) proteins. These molecules have been found to be upregulated in inflammatory contexts such as MS or experimental autoimmune encephalomyelitis (EAE) [59, 60]. In vivo, oligodendroglia expressing MHC class I and class II molecules have been shown to engage CD4 + and CD8 + T cells that have been previously activated [61]. In MS lesions, OPC numbers are reduced, potentially due to these interactions, which may enhance CD8 + T cell-mediated killing of OLs as well as CD4 + T cell-mediated cytotoxicity [61].

Similarly, the ablation of the oligodendroglial TNF receptor 2 (TNFR2) in EAE models exacerbates the immunomodulatory and inflammatory functions of OPCs following stimulation [62]. Collectively, these findings emphasize the multifaceted immunological functions of oligodendroglial cells and underscore their importance in understanding inflammatory myelin-related diseases, such as MS [51].

4. Myelin plasticity

Cerebral plasticity, also referred to as adaptive neural capacity, is defined as the brain's ability to modify its functions in response to processes such as learning. This capacity enables the brain to stabilize these modifications through homeostatic mechanisms and to restore functions following injury. While previous research indicated that this adaptability was exclusively attributable to neurons, recent studies have identified a growing body of evidence indicating the involvement of glial cells, particularly OLs and the myelin they produce [4, 6, 63]. MRI studies in humans have shown that learning can change the profile of white matter [64, 65]. Indeed, it has been demonstrated that the practice of piano, juggling, or the acquisition of reading skills and new languages are all associated with an increase in white matter volume (Figure 4). While this does not provide direct evidence of myelin remodeling, it strongly suggests its involvement. Rodent models have played a particularly instrumental role in elucidating the adaptive properties of myelin. The findings that the adult CNS is not fully myelinated, that individual neurons can display intermittent myelination with unmyelinated segments [41], and that OPCs remain abundant in the adult brain [66], collectively suggest that new myelin can still be generated. A concrete example that clearly suggests myelin could serve as a means of circuit adaptation is the myelination of the neuronal circuit controlling binaural hearing, which enables sound localization [67]. A differential pattern of myelination has been observed between the ipsilateral and contralateral branches of brainstem cochlear nucleus neurons. This differential profile ensures faster signal conduction along the more heavily myelinated contralateral branches, thereby compensating for their longer conduction path. Conduction velocity is contingent upon the degree of myelination, and adaptive myelination exerts a substantial influence on synaptic responses. For instance, it can synchronize the arrival times of electrical signals from different neurons onto a common postsynaptic target, thereby maximizing temporal summation. Consequently, adaptive myelination also addresses specific physiological requirements for the proper functioning of neuronal circuits.

Recent studies using genetic fate mapping have demonstrated that new OLs are indeed generated in the adult brain [66], and that their generation is activity-dependent, much like during development [68, 69]. Remarkably, in vivo two-photon imaging of the adult rodent cerebral cortex has demonstrated unequivocally that new myelin segments are generated during adulthood and that existing segments can both elongate and retract [4, 70, 71].

A seminal study revealed an increase in MBP density in the brain region responsible for motor task learning in rodents [72]. This suggests a correlation between MBP density and motor learning. Another team of researchers demonstrated that the inhibition of new OL generation impairs motor learning [73]. Therefore, the generation of newly formed OLs is necessary for motor learning. It has been shown that motor learning drives the early proliferation of OPCs in the motor cortex and the underlying white matter, with a fraction of them differentiating into myelinating OLs [74]. Recent investigations employing in vivo two-photon imaging have posited that myelin sheaths surrounding learning-activated axons initially retract during the learning process. This retraction leads to the formation of elongated nodes of Ranvier and a new pattern of sparse myelination on these axons [75]. This is believed to result in a transient reduction in conduction velocity during learning tasks [75]. In a subsequent phase, newly generated OLs form new myelin sheaths, preferentially positioned between the retracted existing sheaths [75]. This sequential process establishes a modified pattern of myelination along activated neurons, distinguished by addition of new myelin sheaths [75]. This remodeling has the potential to enhance conduction velocity, thereby facilitating the optimization of signal transmission along consolidated neural circuits following learning.

Memory training in mice also depends on the generation of new OLs [6]. In the water maze paradigm, spatial memory acquisition requires the production of new OLs specifically in the regions activated during training and the subsequent post-training period that supports consolidation [76]. Blocking OPC differentiation during training or consolidation impairs learning and increases latencies in remote recall tasks [76]. Similar results were obtained in contextual fear memory paradigms, in which successful memory formation required OPC proliferation, differentiation into OLs, and subsequent myelin generation [76, 77]. The temporal profile suggests that OPC proliferation and differentiation are crucial during the learning phase, while the formation of new myelin, which emerges weeks after training, plays a more prominent role in memory consolidation than in acute learning [77]. A recent in-depth analysis of oligodendroglial dynamics in spatial working memory tasks confirmed these observations [78]. Memory efficiency was closely linked to both early OPC proliferation and differentiation in stimulated regions and to the overall extent of these processes [78]. Memory consolidation involves subsequent structural remodeling, including the formation of shorter myelin internodes and the insertion of additional sheaths between existing ones [78]. These modifications are expected to increase conduction velocity and mirror the adaptive changes observed in motor learning [78].

Conversely, adverse environmental stimuli have been demonstrated to exert detrimental effects on myelination. In adult mice, social isolation has been demonstrated to induce impairments in social behavior and a reduction in social interest toward conspecifics [79]. In the prefrontal cortex (PFC), it has been reported that there is a global thinning of myelin sheaths, independent of axonal diameter. Reintegration of mice with congeners was able to restore myelin defects as well as social interactions [79]. Remarkably, pharmacological restoration of myelin with clemastine, a muscarinic agonist, was sufficient to rescue social deficits in isolated adult mice [80]. Importantly, social isolation of mice during the juvenile phase also leads to defects similar to those observed following isolation in adulthood [81]. Reintegration with other previously isolated mice does not rescue the defects [81], whereas reintegration with socially housed (non-isolated) animals does [82]. These findings emphasize the remarkable plasticity of myelin and its capacity to modulate cognitive behavior.

Strikingly, myelin plasticity manifests in various ways (Figure 4), including changes in internode length, thickness, and spacing, as well as alterations in the length of the nodes of Ranvier and the width of the periaxonal space, all of which influence neuronal physiology [4, 83].

5. Conclusion

Myelin is no longer regarded as a static structure; rather, it is recognized as a dynamic and adaptive structure capable of remodeling in response to electrical activity and experience. These properties underscore the critical role of myelin in both health and disease. Despite these advances, the molecular mechanisms underlying myelin plasticity remain largely unexplored, opening broad avenues of research that span molecular and cellular biology to membrane biophysics. Advancements in this field are expected to significantly improve our understanding of CNS function and the mechanisms that drive related disorders.

Acknowledgments

We thank the Fondation France Sclérose en Plaques, the Fondation Jérôme Lejeune, and Sorbonne University for their financial support.

The figures were prepared using BioRender, and artificial intelligence was employed to refine and enhance the writing.

Conflict of interest

The authors declare no conflict of interest.

Acronyms and abbreviations