Intrinsic and extrinsic control of oligodendrocyte development

Abstract

Oligodendrocytes (OLs) are the myelinating glia of the central nervous system. Myelin is essential for the rapid propagation of action potentials as well as for metabolic support of axons, and its loss in demyelinating diseases like multiple sclerosis has profound pathological consequences. The many steps in the development of OLs — from the specification of oligodendrocyte precursor cells (OPCs) during embryonic development to their differentiation into OLs that myelinate axons — are under tight regulation. Here we discuss recent advances in understanding how these steps of OL development are controlled intrinsically by transcription factors and chromatin remodeling and extrinsically by signaling molecules and neuronal activity. We also discuss how knowledge of these pathways is now allowing us to take steps toward generating patient-specific OPCs for disease modeling and myelin repair.

Origin and OPC specification

Oligodendrocyte precursor cells (OPCs) arise during development from neuroepithelial progenitor cells in the ventral neural tube (reviewed in Refs. [1–3]). They are specified in response to the secreted morphogen Sonic Hedgehog (Shh) derived from the floor-plate of the neural tube [4]. Before generating OPCs, these same neuroepithelial cells also give rise to motor neurons, and the mechanism of this important switch has recently been shown to involve the hedgehog family member Indian Hedgehog [5], as well as secreted sulfatase-1 [6]. Secreted signals of the Fibroblast Growth Factors family (FGFs) also regulate early OPC development, as loss of either FGF receptor 1 or 2 in ventral neuroepithelial progenitors blocks induction of OPCs [7•, 8•].

Are there functionally different types of OPCs? Recent studies have highlighted the dramatic degree of heterogeneity between astrocytes in different CNS regions [9], but it has been unclear if oligodendrocyte lineage cells might be similarly heterogeneous. In addition to the commonly appreciated ventral OPCs, work by Richardson and colleagues demonstrated that there are multiple waves of OPC specification in the CNS with OPCs being derived from both ventral and dorsal sources [10]. More recently, fate mapping studies showed that ventral and dorsal OPCs myelinate different regions of the brain, raising the possibility of functional differences between them [11]. The observation, however, that different populations of OPCs are functionally redundant and are able to replace each other if one source is ablated [10] would argue against differences. Further work, including a comparison of the gene expression profiles of the different OPC populations, may help elucidate whether they are functionally distinct. Additionally, it will be interesting to understand how dorsally derived OPCs are specified, since dorsally derived Bone Morphogenic Proteins (BMPs) are typically thought to compete with Shh to inhibit OPC specification.

Recently it has been found that OPCs are also generated in the adult brain from neural precursor cells in the subventricular zone [12]. Their specification may rely on the same signals as during development, including Shh and FGF-2 [13, 14]. Other work, however, suggests that additional signals may be required, including epidermal growth factor (EGF) and pigment epithelium-derived factor (PEDF) [15, 16]. A major challenge in demyelinating diseases like multiple sclerosis is the fact that, although OPCs are prevalent in demyelinated lesions, they often fail to differentiate into OLs and remyelinate. Are the signals that promote differentiation of adult OPCs different than the ones used during development, and if so, can we harness that knowledge to promote remyelination? One recent study by ffrench-Constant, Franklin and colleagues generated gene expression profiles for OL differentiation following focal demyelination [17]. Intriguingly, the authors found that signaling through the retinoid X receptor gamma is important for differentiation of adult OPCs and remyelination, despite not seeming to play a role in developmental myelination. In another elegant study, it was found that the systemic milieu of young mice promotes OL differentiation and remyelination when introduced into the aged mouse by heterochronic parabiosis (connecting the blood stream between the young and old mouse), by unknown mechanisms [18••]. These studies are of particular relevance for demyelinating disease, as understanding the signaling pathways that lead to the generation of new OPCs and their differentiation may prove useful for stimulating remyelination in patients [19].

How do intrinsic factors regulate the differentiation of OPCs into OLs?

OPCs undergo a defined series of steps to differentiate into a premyelinating OL, extend processes that wrap and compact around axons, and finally turn on genes required for myelin maintenance and the (still emerging) physiological functions of mature myelin [20] (Figure 1). There are many steps in this process that must be regulated in time and space, and thus there is a great need to uncover how the many transcriptional regulators in OPCs/OLs collaborate to precisely coordinate these cellular changes [21]. This involves both activation of genes to promote differentiation, and repression of genes that prevent differentiation [22]. How do OPCs/OLs control this very specific and coordinated timing of gene expression? Over the past decade it has become clear that chromatin remodeling is a key process that regulates OL development [22–24]. There are two major ways that chromatin is remodeled to change gene expression: Firstly, covalent modifications of histones such as deacetylation by HDACs to silence genes, or acetylation to activate them, and secondly, ATP-dependent remodeling by SWI/SNF enzymes such as Smarca4/Brg1, that control the position of nucleosomes and thus increase the accessibility of specific genes.

Figure 1.

Regulation of oligodendrocyte development. Top: Fluorescence micrographs of primary cultured oligodendrocyte precursor cells (OPCs), premyelinating oligodendrocytes (early OLs), and OLs myelinating unlabeled CNS axons. NG2 chondroitin sulfate proteoglycan (green) labels OPCs, while myelin basic protein (MBP, red) labels differentiated OLs and is found in compact myelin. Bottom: Summary of extrinsic signals that regulate OPC specification and proliferation, and intrinsic factors affecting OL differentiation, as described in the text.

Work by Casaccia and colleagues first established that histone deacetylation by HDACs plays important roles in OL development [25]. Recently Lu and colleagues used genetic ablation to show that HDAC1 and HDAC2 are required for OL differentiation, and control the Wnt signaling pathway, a known inhibitor of OL differentiation [26]. Similarly, in the PNS, HDAC1 and HDAC2 are necessary to regulate transcription by Schwann cells [27, 28], suggesting that this is a conserved pathway in all myelinating glia. Intriguingly, extrinsic factors that regulate OL differentiation appear to do so, at least in part, by modifying histone acetylation. Shh, which promotes OL differentiation, induces histone deacetylation by HDACs; in contrast, inhibitory BMP4 appears to block deacetylation [29]. Bioinformatics approaches to identify sequential waves of co-regulated genes during OL differentiation found that genes with similar functions are co-regulated to turn on or off at the same time; importantly, putative inhibitory genes require HDAC activity to turn off at the onset of differentiation [21]. Taken together it seems that the major role of HDACs is to repress expression of genes that normally block OPC differentiation, thereby ‘releasing the brakes’ and allowing differentiation to proceed [21].

Once inhibition is relieved, what is the mechanism that drives transcription of genes required for differentiation? It has long been known that the OL-specific transcription factors Olig1 and Olig2 play an essential role in OL specification and differentiation [30, 31]. Until recently it was unknown how Olig1 and Olig2 actually function to promote differentiation. Lu and colleagues demonstrated that Olig2 binds to enhancers of OL-specific genes and recruits the SWI/SNF chromatin remodeling protein Smarca4/Brg1 [32••]. This suggests a model of Olig2 function in which it promotes ATP-dependent chromatin remodeling to make the promoters of OL genes accessible for transcription. Interestingly, their genome-wide analysis of Smarca4/Brg1 and Olig2 co-occupancy on chromatin suggests novel transcription factors and other genes that may be required for OL differentiation and myelination. Testing the roles of these genes, as well as determining whether chromatin remodeling by HDACs and Smarca4/Brg1 is important for remyelination, are important questions for future work [33].

It is likely that a number of transcription factors and other regulatory molecules collaborate to regulate OL differentiation. Both Myelin Gene Regulatory Factor (MYRF/gm98) and Zfp191 appear to be master regulators of myelin gene expression [34, 35], and MYRF is required in adult OLs to maintain OL identity [36•]. The precise mechanisms of how these genes regulate OL differentiation remain to be elucidated. An intriguing new study indicates that the OL transcription factor Sox10 directly interacts with the Mediator complex to recruit RNA polymerase II during OL differentiation [37]. Recent work suggests that master transcriptional regulators collaborate with the Mediator complex to recruit RNA polymerase II to ‘super enhancer’ domains, to regulate cell identity [38]. It will be interesting to know whether this is a general feature of OL differentiation.

Work in the past few years from multiple groups has also shown the importance of micro-RNAs (miRNAs) in OL development, myelination, and remyelination [39]. In addition to miRNAs, the list of functional types of noncoding RNA continues to grow, including long noncoding RNAs [40]. Increasingly it is appreciated that many of these play important roles in regulating gene expression directly or by acting as ‘sponges’ to compete with miRNAs or endogenous RNA [41, 42]. With better technologies enabling the detection and quantification of these other RNA species (e.g. RNA-seq whole transcriptome sequencing technologies), it will be interesting to see whether they are expressed by OPCs/OLs and play unique roles in their development.

Extrinsic control of OL development: the role of neuronal activity

It has long been appreciated that neuronal activity is likely to be important for the development of OLs [43]. Within the past decade, work by Bergles and subsequent studies from multiple groups have demonstrated that OPCs receive functional synapses from neurons [44, 45], but the precise role of these synapses in vivo has yet to be determined. One attractive possibility is that OPCs sense neuronal activity to regulate their capacity to divide, differentiate, and myelinate [46]. Neuronal activity can influence OL differentiation and myelination in vitro via secreted signals, including glutamate and ATP [47•]. Confirmation of these results in vivo, perhaps using an optogenetic approach, would be an exciting step forward, and much of the groundwork has been laid.

How do electrically active neurons signal to OPCs? In the case of excitatory neurons, the answer might be the neurotransmitter glutamate. Although ablating NMDA receptors from OPCs in vivo has no effect on OPC proliferation, myelination, or responses to white matter injury, this causes an upregulation of AMPA receptors that could also respond to neuronally derived glutamate [48, 49]. Indeed, it appears that AMPA receptors, as well as purinergic receptors that respond to ATP, help OPCs to respond to neuronal activity in vitro [50], but whether this also occurs in vivo remains to be established.

Several recent in vivo studies have provided intriguing evidence that neuronal activity regulates OL development and myelination. For instance, the formation of neuron-OPC synapses in the mouse barrel cortex is regulated by sensory input [51]. Strikingly, social experience also powerfully regulates myelination in the mouse prefrontal cortex, and in the adult this may be regulated by chromatin remodeling [52••, 53••]. As loss of the ErbB3 receptor leads to the same myelination defects as sensory deprivation, neuregulin-ErbB signaling likely plays a role in this form of myelin plasticity [52••]. This is especially interesting since neuregulin-ErbB is a master regulator of myelination in the PNS, but until now has appeared to be largely dispensable for CNS myelination. Another study showed that in vivo electrical stimulation of CNS neurons of the corticospinal tract with high-frequency stimulation induces proliferation of OPCs in the adult rat [54]. Moving forward, it will be important to determine whether these OPCs myelinate axons, and whether physiological (lower frequency) activity in neurons is sufficient to regulate OPC proliferation or differentiation. Clearly, a major goal for the field is to determine which stages of OPC/OL development and myelination are controlled by normal neuronal activity in vivo, and to elucidate the signaling pathways that regulate this developmentally and in disease. Whether OPCs respond to neural activity for functions other than regulating myelination remains an intriguing possibility [3, 45].

While much focus has been given to how neuronal activity may affect OPCs during development, it is also important to consider OPCs in the adult CNS. Given that the bulk of myelination occurs early after birth, it is perhaps surprising that OPCs remain abundant in the adult brain. An elegant recent study by Young, Richardson and colleagues used lineage tracing and EdU labeling (of dividing cells) to measure the extent of OPC proliferation and differentiation into OLs in the adult mouse brain [55••]. This work establishes some general features of all OPCs in the adult CNS: they continue to divide throughout adulthood, and form new myelinating OLs with progressively more and shorter myelin internodes. The continual addition of new myelin suggests that, in contrast to long held views, there is normally much more myelin turnover in the adult brain than has been thought. The precise mechanism of this (turnover of OLs by apoptosis, turnover of individual internodes, or addition of new internodes between existing ones) and signals that regulate it remain to be determined. A provocative possibility is that myelin remodeling is involved in learning processes [46].

Generation of OPCs from stem cells and direct lineage conversion

How can we translate what we are learning about OL development into therapies for demyelinating diseases? A major goal in the field is to be able to generate patient-specific myelinating cells, both for disease modeling studies and eventually to be used therapeutically to regenerate myelin [56, 57]. Ideally these cells would be proliferative and maintain their ability to myelinate, so generating OPCs is preferred to generating mature OLs, which quickly lose their capacity to myelinate [58]. Our increasing understanding of the signals that regulate specification and development of OPCs has allowed major breakthroughs toward this goal in the past few years.

A useful assay has been developed utilizing the hypomyelinating shiverer mouse that lacks expression of Myelin Basic Protein (MBP). Injection of experimentally derived wild-type OPCs or progenitors into shiverer mice allows for unambiguous identification of transplanted cells, which uniquely express MBP after differentiating into OLs [59]. Importantly, these wild-type OLs can compete with endogenous (MBP-lacking) OLs and generate myelin, enabling researchers to determine whether generated OPCs are functional in vivo.

In one approach, human fetal OPCs are purified based on specific surface expressed epitopes; these can be expanded in vitro, and engraft and widely myelinate within the mouse CNS after transplantation [60]. A second approach relies on banked and proliferated human neural stem cells, which can also myelinate the mouse brain [61] and, strikingly, safely engraft in the human brain of patients [62•]. Toward the goal of generating patient-specific OPCs, several recent papers have built on work in the stem cell field to generate OPCs from pluripotent stem cells. Both rodent epiblast stem cells and human induced pluripotent stem cells (iPSCs) can myelinate in vivo and even appear to dramatically increase the survival of hypomyelinated shiverer animals [63, 64•]. A potentially more rapid method of generating patient-specific OPCs is to reprogram already differentiated cells, a technique known as direct lineage conversion [57]. In this approach, fibroblasts or other differentiated cells are directly converted into another cell type by forced expression of transcription factors or other regulatory molecules. Two recent papers demonstrated that rodent fibroblasts can be converted to the OL lineage by expression of three transcription factors: Olig2, Sox10, and either Nkx6.2 or Zfp536 [65•, 66•]. Important next steps will be to translate these findings to human cells, and to test and optimize their ability to restore function to the de/dysmyelinated CNS.

Summary and outlook

It is an exciting time in the OL field, and recent advances have provided a great deal of mechanistic insight into how OLs develop in response to intrinsic and extrinsic signals. However, there are still many fundamental questions remaining about the functions of OPCs and OLs in the CNS, and exactly how OPCs differentiate into OLs. A clear goal of future work is to test, in vivo, the many predictions made from in vitro studies. This includes further work to elucidate how sequential stages of OL differentiation and myelination are regulated transcriptionally, understanding the roles of OPC-neuron synapses and other OPC functions in the adult brain, and determining how neuronal activity affects OL development and myelination. In addition, a major translational goal is to refine current methods to generate and transplant OPCs into patients, with an ultimate goal of regenerating lost myelin in the human CNS.