In vitro modeling of central nervous system myelination and remyelination

Abstract

This review aims to summarize the current techniques to study myelination and remyelination in culture systems. We attempt to put these into historical context, and to identify the strengths and weaknesses of each approach, which vary depending on the experimental question to be tested. We discuss the difficulty and importance of quantification of myelination and in particular remyelination. Finally, we provide our predictions of how these techniques will and should develop in the future. © 2011 Wiley Periodicals, Inc.

INTRODUCTION

The appearance of myelin, and especially myelination of central nervous system (CNS) axons by oligodendrocytes, was a crucial step in vertebrate evolution. The electrically insulating properties of myelin allow rapid, high-fidelity propagation of action potentials along the axon without a corresponding increase in axonal diameter, resulting in the development of significantly increased complexity and size of the organism. Sensory, motor, and cognitive ability is severely compromised following damage to myelin in disease states provoked by either genetic mutation, as with the leukodystrophies, or by autoimmune attack, as in multiple sclerosis, the most prevalent CNS demyelinating disease. Remyelination of CNS axons can occur, and this has been shown to restore nerve function, yet it is inefficient and insufficient in most multiple sclerosis patients. Understanding how oligodendrocytes initially produce myelin during development and remyelinate axons in the diseased CNS is of significant clinical interest, as it opens up avenues for novel remyelinating therapies, using either endogenous or exogenous sources of oligodendroglia. In vitro culture models of myelination and remyelination can help dissect out these processes.

Tissue culture approaches have been used for the study of myelin since the mid-1950s. Although these early studies using organotypic and aggregate cultures allowed investigators to improve their understanding of myelin ultrastructure, it was the development of culture methods allowing for the isolation of oligodendrocytes over three decades ago (McCarthy and de Vellis,1980) that led to a more refined understanding of how oligodendrocytes mature and produce myelin. A key breakthrough was the identification of stage-specific markers (reviewed by Baumann and Pham-Dinh,2001), which allowed for the precise classification of oligodendroglial differentiation state. The expression of these markers occurs on schedule in oligodendroglial cultures, as in vivo, suggesting this is intrinsic to the cells. Morphological changes associated with those seen in vivo during myelination are also present in these cultures. When cultured in the absence of axons, differentiated oligodendrocytes are capable of synthesizing membrane sheets whose organization and composition closely resembles that of myelin (Szuchet et al.,1986). Oligodendrocyte morphology can be scored as immature (simple processes), more mature (complex processes), or mature and forming myelin membrane sheets (Colognato et al.,2007; Olsen and ffrench-Constant,2005). If morphology must be quantified, process complexity, which can be measured by Sholl analysis (Rajasekharan et al.,2009) or fractal dimension (Behar,2001) can be used to measure the extent of morphological shape changes correlated with maturity and myelination, but only prior to membrane sheet formation when processes merge and the shape is no longer a fractal.

However, these cultures do not model all aspects of myelination effectively. Establishment of axo-glial contact results in vastly increased synthesis of myelin membrane by oligodendrocytes. Each myelinating oligodendrocyte in vivo can produce approximately 500 times the area of myelin membrane than an oligodendrocyte in culture (Pfeiffer et al.,1993). This is equivalent to comparing the size of a shirt button to the size of a wheel of a small family car. If cultured oligodendrocytes produced a similar amount of membrane in the absence of neurons, a 13-mm circular coverslip could be completely covered by membrane sheets produced by just 60 oligodendrocytes! While oligodendrocytes are capable of wrapping inert materials such as carbon (Althaus et al.,1984), glass and vicryl microfibers (Howe,2006) and fixed neurons (Rosenberg et al.,2008), the extent of wrapping is considerably decreased compared with that seen with living neurons. Oligodendroglial development, myelination, and remyelination are known to regulated by secreted factors, (Allamargot et al.,2001; Ishibashi et al.,2006; Jean et al.,2002; Piaton et al.,2011), contact-mediated signals (Camara et al.,2009; Charles et al.,2000; Zonta et al.,2008), and electrical activity (Demerens et al.,1996; Stevens et al.,2002) which all require the presence of neurons (and, in some cases, astrocytes). The most advantageous tissue culture approaches to improve understanding of basic CNS myelin biology and to develop remyelinating therapies are those in which these interactions between axons and oligodendrocytes are maintained in vitro, the focus of this review.

SUMMARY AND COMPARISON OF CURRENT TECHNIQUES USED TO MODEL MYELINATION AND REMYELINATION (Fig. 1)

MYELINATING CO-CULTURES

Although cultures of purified or enriched oligodendrocyte lineage cells can be used to study aspects of oligodendroglial cell biology, the existence of multiple reciprocal signaling interactions between oligodendrocytes and neurons prior to and during myelination implies that more meaningful insights can be gained through the use of myelinating co-culture systems. Myelinating co-culture systems are advantageous because manipulation of gene expression and signalling in vitro is relatively simple, and the cultures are easy to image and analyze.

Figure 1. Summary of culture models of myelination and remyelination. Dissociated cultures: Cerebral hemispheres are dissected from whole brains, dissociated, and plated out on coverslips, where oligodendrocytes myelinate axons. Oligodendrocyte-DRG co-cultures: Cerebral hemispheres are dissected from whole brains and oligodendrocyte precursor cells extracted, usually by immunopanning or differential adhesion. These are then added to cultures of DRG neurons obtained from embryonic spinal cord and myelination occurs in around 2 weeks. Oligodendrocyte-CNS explant co-cultures: Embryonic spinal cord segments are either grown on Matrigel coverslips preseeded with oligodendrocyte precursor cells, or on a bed of astrocytes, in which case OPCs migrate out of the explant to myelinate axons. Slice cultures: Newborn slices of rodent brain or spinal cord are cultured on organotypic membranes to study myelination. Treatment with a myelin toxin (such as LPC) or anti-myelin antibodies (and complement) causes demyelination, which is followed by remyelination. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Several approaches can be used to prepare myelinating cultures. The most common approaches include dissociating embryonic CNS tissue and then maintaining that mixed cell population as a two- or three-dimensional culture, or culturing dissociated neurons or explants of nervous tissue from one source and later adding oligodendrocytes from a second source. Each culture method, as well as their advantages and disadvantages will be described below.

Dissociated CNS cultures

The first reports of myelination in two-dimensional dissociated culture describe experiments in which immature CNS tissue is broken up by proteolysis, and intercellular connections are reestablished in the resultant mixed cell population. In the earliest of these, myelination was only observed after very long periods in culture. Initial CNS myelin formation in preparations derived from 11-day-old chick spinal cord was observed after 24 days in vitro, with mature compact myelin visible after 30 days (Kim,1972), while similar preparations from embryonic day 13–14 (E13-14) mouse spinal cord/brain stem showed evidence of myelination after 14 days (Bornstein,1972). A significant refinement to CNS culture methods was the use of polylysine instead of collagen as a substrate. This allowed the majority of cells to rapidly adhere (Yavin and Yavin,1974), and resulted in the appearance of myelin in cultures prepared from E16 rat cortices after as few as 8 days in vitro (Yavin and Yavin,1977), a time course that is far more consistent with rat cortical developmental myelination in vivo (Hamano et al.,1996).

The method used to prepare dissociated CNS cultures has been further refined more recently. While older protocols called for high concentrations of serum (10–20%), the development of defined media allowed serum concentrations to be reduced by an order of magnitude (Bottenstein and Sato,1979), allowing far greater control over experimental conditions. Cultures prepared from dissociated E15 mouse cortices revealed that all oligodendrocyte processes myelinate only axons, that wrapping by each process of a given oligodendrocyte is synchronized, that myelination is dependent on axon density, and that the number of myelinating processes per oligodendrocyte in vitro is consistent with the number of wraps formed per cell in the corpus callosum in vivo (Lubetzki et al.,1993). Very similar findings were reported later in a different co-culture system (Watkins et al.,2008) (see later). A similar approach was used to prepare myelinating cultures from dissociated E15–17 rat cerebellum, hippocampus, and spinal cord, allowing for in vitro analysis of myelination of neurons from specific brain regions. This study also demonstrated that use of Matrigel rather than polylysine as a substrate accelerates myelination (Svenningsen et al.,2003). Further validation confirmed that myelination in these cultures (in this case, from E13.5 mouse spinal cord) is consistent between experiments, and that myelin thickness relative to axonal diameter (G-ratio) is consistent with that observed in vivo (Thomson et al.,2006).

Using dissociated CNS cultures to study myelination has two significant advantages. First, it can be argued that these cultures faithfully recreate CNS myelination, as the neuron and glial populations present normally encounter each other in vivo. In addition, the method allows for the preparation of a sufficiently large quantity of cultures for biochemical analyses from a relatively small number of mouse or rat embryos (Thomson et al.,2008). The use of a common source of neurons and oligodendrocytes, however, represents a limitation of this technique, as gene expression cannot be manipulated independently in each population of cells.

Oligodendrocyte-dorsal root ganglion neuron co-cultures

A second myelinating co-culture system that has been in use for decades is comprised of dorsal root ganglion (DRG) neurons maintained in culture in the presence of nerve growth factor (NGF) and an anti-mitotic agent (to kill Schwann cell progenitors and other contaminating cells), to which oligodendrocytes (prepared independently) are added following growth factor withdrawal. In the initial description of this method, oligodendrocytes were added by overlaying short sections of optic nerve obtained from 1- to 2-week-old postnatal rats onto 2-week-old DRG neurons, and co-cultures were maintained in a highly-enriched medium containing 25% human placental serum and 10% chick embryo extract. Sudan black staining and electron microscopy were used to demonstrate that, after 5 weeks, oligodendrocytes had wrapped axons with compact myelin (Wood et al.,1980). Later studies described further refinements of this culture method. Serum concentrations were reduced (Wood and Bunge,1986b; Wood and Williams,1984) and eventually serum-free defined media were developed that could promote myelin formation in vitro (Rosen et al.,1989), allowing for more precise assessment of what factors influence myelination. Although myelination was only observed 4 or 5 weeks following the addition of oligodendroglia to neurons in these early studies, current protocols in which medium composition and axonal and oligodendrocyte density have been optimized allow dense myelination to be obtained after only 10–14 days after the addition of oligodendrocyte precursor cells (OPCs) (Chan et al.,2004; Huang et al.,2011; Wang et al.,2007).

In studies published during the early 80s, the reported survival of neurons in dissociated CNS cultures was highly inconsistent from study to study, leading to highly variable oligodendroglial proliferation and differentiation in vitro. The relative robustness of DRG neurons in the oligodendrocyte-DRG co-culture system, using distinct cells from clear sources, allowed experiments to be carried out that led to many advances in the understanding of how both secreted and contact-mediated axonal signals influence oligodendroglial development. These include the demonstration that axons promote the proliferation of embryonic (Wood and Williams,1984) and adult (Wood and Bunge,1986a) OPCs, and that OPCs from adult CNS can differentiate into myelin-forming oligodendrocytes (Wood and Bunge,1986b). Later studies suggested that the OPC mitogens platelet-derived growth factor (PDGF) and fibroblast growth factor-2 (FGF-2) could inhibit myelination, while soluble neuregulin increased myelin formation (Wang et al.,2007). Oligodendrocyte-DRG neuron cultures have also been used to demonstrate how axo-glial contact maintains oligodendroglial responsiveness to neurotransmitters, and the activation of multiple neurotransmitter-activated signal transduction pathways (He et al.,1996), while xenocultures of mouse oligodendrocytes and rat DRG neurons were used to characterize changes in integrin expression during myelination, with species-specific antibodies being used to distinguish between glial and axonal integrins (Shaw et al.,1996).

Oligodendrocyte-DRG neuron co-cultures have also been used to study the effect of spatial constraints on oligodendroglial development. Provocative findings by Rosenberg and colleagues suggest that packing constraints are critical for the regulation of oligodendrocyte differentiation in the presence of DRG axons, and that differentiation requires that OPCs be sufficiently spatially constrained. While this may be achieved by the presence of a sufficient OPC density, it can also be achieved using a similar density of Schwann cells or cell-sized (but not larger or smaller) polystyrene beads conjugated to an antibody that recognizes axons. Especially surprising is their further observation that under high packing density conditions, oligodendrocytes can initiate the wrapping of axons that have been fixed with paraformaldehyde, suggesting that active signalling from axons is not required to promote myelination (Rosenberg et al.,2008).

This final study exemplifies a major advantage of the oligodendrocyte-DRG co-culture system: its flexibility. Oligodendrocytes obtained from different species, CNS regions, tissues of different ages, or prepared by different methods (from tissue explants, shaking off from mixed glial cultures, immunopanning, magnetic or fluorescent-activated cell sorting) can all be used to myelinate DRG neurons. Because they are first cultured separately, either neurons or oligodendrocytes can be manipulated independently of each other prior to co-culture, and other cell populations (or non-cellular material) can be introduced as well. Other advantages include the use of defined media, relatively modest numbers of animals required per experimental replicate, the ease with which these cultures can be imaged, and the ability of co-cultures to survive for up to 8 weeks following the addition of oligodendrocytes.

A related weakness of this technique is that considerable time is needed to complete each experiment because DRG neurons must be maintained in culture for 3 weeks to survive independently of NGF before addition of oligodendrocytes, as NGF was shown to inhibit myelin formation by oligodendrocytes (Chan et al.,2004). These cultures then must be maintained for a further 2 weeks to allow myelination to occur. Perhaps the most significant disadvantage of this culture method is that DRG neurons may not perceived to be “true” CNS neurons as their cell bodies are located outside the CNS and they form connections with both CNS and PNS neurons. Consequently, the applicability of any findings using this culture method to CNS myelination in vivo can be called into question. While no thorough comparative study has been carried out, previous findings have demonstrated that the lengths of myelin internodes produced by oligodendrocytes co-cultured with DRG neurons (Camara et al.,2009) bear a striking similarity to those measured for cortical oligodendrocytes in vivo (Murtie et al.,2007), suggesting that, by one metric at least, oligodendrocytes behave similarly in both environments (Fig. 2). In addition, as DRGs send one projection into the CNS and one into the PNS, it can also be argued that the receptivity of DRG neurons for myelination by both Schwann cells and oligodendrocytes makes these cells ideal for studies comparing signals influencing central and peripheral myelination.

Figure 2. The length of myelinated internodes by oligodendrocytes in co-culture (A) with DRG neurons is comparable to that measured for cortical internodes in vivo (B). (Reproduced with permission from Camara et al., Journal of Cell Biology,2009, 185, 699–712 © The Rockefeller University Press and Murtie et al., Journal of Neuroscience Research,2007, 85, 2080–2086 © The Society of Neuroscience).

Oligodendrocyte-CNS neuron co-cultures

To address the major weakness of DRG neuron co-cultures, culture systems in which oligodendrocytes and CNS neurons are derived from independent sources have been established. In the first of these, rat retinal ganglion cell (RGC) neurons were purified by immunopanning from retinae of P5 animals, plated at high density on chamber slides for 2 days to allow neurons to reaggregate, and then distributed on PDL-laminin-coated glass coverslips. After approximately 10 days in culture to allow for axon growth, purified OPCs were added to the RGC cultures. The authors used this culture system to demonstrate that in the presence of γ-secretase inhibitor DAPT, myelination could be observed after as little as 3 days of co-culture, compared with 6 days in controls (Watkins et al.,2008). In addition to its use of CNS neurons and the rapidity of myelination, this method shares all the advantages of manipulability of DRG-neuron co-cultures. A significant drawback to this method, however, is the technical difficulty of preparing these cells. The preparation of 24 coverslips of reaggregated neurons required approximately 2.2 × 10⁶ RGCs from three litters of rat pups, or six litters of mouse pups.

A second approach used to prepare CNS neuron-oligodendrocyte co-cultures employs explants of E14 rat spinal cord cultured on Matrigel-coated coverslips pre-seeded with OPCs. Neurite outgrowth occurred principally for 3–4 days after addition of the explant, with myelination being observable after 15 days (Chen et al.,2010). This method has the significant disadvantage of requiring the use of fluorescently (or otherwise)-labeled oligodendrocytes to discriminate between oligodendrocytes originating from within and without the explant.

Aggregate cultures

Aggregate cultures were the earliest method used to successfully study large numbers of CNS neurons in vitro. They were developed in the early 1970s to circumvent the technical challenge of poor neuronal attachment to early tissue culture substrates. Following enzymatic digestion and mechanical dissociation of foetal or neonatal rat CNS brain tissue, cultures are incubated in an oxygenated, CO₂-enriched environment under constant rotation for at least several weeks. Under these conditions, cells were shown to form aggregates, with cellular organization (DeLong,1970) and “biochemical differentiation” (a measure of the activity of three neurotransmitter-metabolising enzymes) (Seeds and Vatter,1971) broadly resembling what was observed in vivo. Electron microscopy further revealed the presence of synapses, which increased in number with culture time, and the presence of myelin after 5 weeks in vitro (Seeds and Vatter,1971). A further study by Matthieu et al. (1978) revealed that the concentration of myelin components myelin basic protein (MBP), CNPase and sulfatide increased in aggregates at times consistent with their corresponding increases in vivo, though myelinated axons were observed relatively late. As with other culture techniques, a key refinement was the development of serum-free chemically defined medium (Honegger etal.,1979), which allowed aggregate cultures to be used to further characterise myelination in vitro (Trapp et al.,1982), and identify factors that influence myelination (Almazan et al.,1985; Tosic et al.,1992)}. A further development was the preparation of aggregate cultures from mouse (Berglund et al.,2004), which allowed this technique to be used in studies using tissue from transgenic mouse lines (Jackson et al.,2004).

Unlike “two-dimensional” myelinating cultures, aggregate cultures can be used to study demyelination and remyelination. Exposure of 5-week-old rat brain aggregate cultures to an antibody against MOG and complement resulted in decreased MBP expression (Kerlero de Rosbo et al.,1990) and loss of myelin sheaths (Loughlin et al.,1997). Approximately a week following removal of MOG antibody and complement, expression levels of MBP recover to control levels (Matthieu et al.,1992), and an increase in the number of myelinated axons was observed (Diemel et al.,2004). Recently, rat and mouse aggregate cultures have been used to study remyelination following demyelination with the myelin toxin lysophosphatidylcholine (LPC) (Vereyken et al.,2009). Aggregate cultures have also been employed to compare cellular responses to anti-MOG and complement, LPC, and a third model of demyelinating injury, interferon-γ treatment with lipopolysaccharaide (LPS) (Defaux et al.,2010). It has not been demonstrated in these studies, however, that myelin formed following demyelination in this system is thinner or displays decreased internodal length consistent, two hallmarks of remyelination invivo (Blakemore,1981; Bunge et al.,1961; Perier and Gregoire,1965), rendering it difficult to determine whether myelin formed following demyelination in these cultures more closely resembles remyelination or ongoing myelination. However, the response of microglia/macrophages and increased astrogliosis broadly resembles that reported in vivo following demyelination (Defaux et al.,2010). In addition, it has been reported that OPCs are continuously present in the cultures (Diemel et al.,2004), and a proliferative response (though, surprisingly, no increase in PDGFαR immunoreactivity) is observed following demyelination (Vereyken et al.,2009), consistent with endogenous progenitor cells being the likely source of remyelinating oligodendrocytes in these cultures as has been reported in vivo (reviewed by Keirstead and Blakemore,1999; Woodruff et al.,2004).

While the principal advantages of aggregate cultures are that they are fully CNS-derived and their structure is three-dimensional, making them ideal for studies of demyelination and remyelination, the main challenge to use of this technique is practicality, as they require specialized equipment that renders simultaneous assessment of a large number of experimental conditions impractical for most laboratories.

SLICES

A disadvantage of the in vitro cell culture systems described earlier is that the true three-dimensional structure of the tissue is absent, and the cells are asked to behave “normally” in very contrived environments. An alternative is to use an ex vivo slice culture model, where the three dimensional structure of the brain is retained within the slice. Initially, this slice technique was developed as a tool to record neurophysiology in acute brain slices, but longer-term culture was soon seen to be desirable.

Ex-vivo culture of brain slices is a technique which dates back to 1941 when Levi and Meyer (1941) attempted culture of embryonic chick rhomboencephalon, but long-term cultures were not successful. In the 1950s, the technique was developed further with use of plasma drops to maintain the cultures. The original description of myelin formation in culture of mammalian CNS was made in 1956 and was visualized by its characteristic refractive pattern on light microscopy (Hild,1956). Longer term kitten and rat cerebellar cultures grown on rat collagen confirmed this (Bornstein,1958) and the presence of myelin was identified using Luxol Fast Blue and Sudan Black staining (Bornstein and Murray,1958). Culture of spinal cord slices occurred a little later with successful culture of transverse sections of fetal rodent and human spinal cords in 1965 (Peterson et al.,1965), primarily to study nerve outgrowth and synapses electrophysiologically, or neuronal location (Sobkowicz et al.,1968). At this stage, biologists were unclear about which cells actually formed myelin in the CNS (in vivo) due to technical limitations in identifying specific cells. However, electron micrographs using these rat and mouse cerebellar cultures enabled the visualisation of the initial wrapping of axons by glial cells and later compaction of myelin sheaths (Field et al.,1969; Kim,1971; Ross et al.,1962). Later, the advent of immunohistochemistry allowed oligodendrocytes to be confirmed as the myelinating cells of the CNS.

The success of longer-term slice cultures also depended on technical advances. Initial slices were hand cut into cubes of tissue, whereas later these were refined to more uniform slices of 200–500 μm cut with a tissue chopper or vibrating microtome. Roller cultures (Gahwiler,1984) were used to obtain sufficient aeration and as the cultures thinned to ∼50 μm rapidly, analysis using microscopy was easier. However, these cultures often failed to properly adhere, and the system had to be disassembled for feeding and examination under the microscope. Slices were also grown in a drop of nutrient medium between two collagen covered coverslips in Maximow assemblies, but this also needed to be dissembled for feeding and processing and was expensive and laborious. Long-term culture of spinal cord for many months was possible when the cord was cut into longitudinal sections (rather than transverse) and grown on small moulded fluoroplastic dishes coated with collagen (Bunge and Wood,1973). Currently, slices are grown on semiporous membranes (e.g. Millipore organotypic inserts), which cause less tissue thinning and allow for much simpler maintenance and processing (Stoppini et al.,1991). When cultured in this manner, slices of mouse cerebellum, brainstem, cerebral hemisphere, and spinal cord can all be maintained in culture for at least 4 weeks (Zhang et al.,2011), though cerebellar cultures can be viable for significantly longer (Jarjour et al.,2008). The change to the routine use of mouse slices has enabled study of myelination in transgenic mice and those with naturally occurring mutations.

Mutant/Transgenic Mice

As mice with natural mutations and hypomyelinating phenotypes became available, slice cultures from newborn mice were used to study the cell biology of these mutations. For example, cerebellar slices from quaking (qk) and jimpy (jp^msd (myelin synthesis deficiency) and jp) mice were examined using semi-thin and ultra-thin sections to assess whether myelin defects were intrinsic to the myelin or were caused, or at least influenced, by a circulating deleterious substance (Billings-Gagliardi etal.,1980). The slice technique also allows the study of myelination in cerebellar cultures using transgenic mice which otherwise die before myelination occurs. For example, transgenic mice null for Netrin-1 or DCC die within hours of birth (for non-neurological reasons) but by study of cerebellar slices from these mice, it was seen that myelination occurred with abnormalities of the paranodal loops (Jarjour et al.,2008). Direct visualization of myelinating cells in slice cultures can be achieved using transgenic mice expressing a fluorescent protein under an oligodendrocyte promoter (e.g. PLP-GFP) (Harrer et al.,2009). Delivery of genes, e.g. GFP or sh/mirRNAs directly to cells within slices to manipulate cell biology is also possible using lentiviruses (Zhang et al.,2011) or Semiliki Forest Virus (Haber et al.,2009).

Transplantation

Manipulation of myelination in the slice system can also be achieved using cells transplanted into the slice. Cerebellar slice cultures, exposed to cytosine arabinoside to prevent oligodendrocyte differentiation, did not myelinate until another slice (treated with kainic acid to eliminate Purkinje cell neurons), was transplanted on top. The slices then myelinated well, suggesting that myelinating cells migrated out of the upper slice into the lower slice (Seil and Blank,1981). Cultures of shiverer mouse cerebellum, with normal optic nerve in direct contact (containing normal oligodendroglial cells) also make normal myelin both on electron microscopy and by immunohistochemistry to MBP (Billings-Gagliardi et al.,1984). This suggests that normal oligodendrocytes migrated from the optic nerve into the shiverer cerebellum and this is sufficient to form normal myelin sheaths around shiverer axons. Transplantation of purified rat oligodendroglial cells (McCarthy and de Vellis,1980) into mouse cerebellar explants treated with cytosine arabinoside, thus preventing endogenous myelination, achieved some successful myelination presumed to be myelination of mouse axons by rat oligodendroglial cells (Nishimura et al.,1985). Purified OPCs, transduced with lentiviral construct containing GFP, were added to mouse cerebellar slices and formed GFP+ myelin sheaths (Zhang et al.,2011), opening the way for genetic manipulation of OPCs and studying their behavior in myelination, demyelination, and remyelination.

Demyelination

Even as early as 1959, cerebellar slices were successfully demyelinated using serum from rabbits with experimental allergic encephalomyelitis, in the presence of complement (Bornstein and Appel,1959), although this was more as a test of demyelination capacity of antibodies in the serum rather than studying the process of demyelination itself. Since then, various methods of demyelination have been developed. Addition of rabbit sera raised against MBP, GalC or white matter to well myelinated embryonic mouse slice cultures caused demyelination (Raine et al.,1981). Addition of anti-white matter serum with complement also demyelinated spinal cord slices, and IGF-1 both inhibited this and accelerated remyelination as defined by a return of CNPase activity (Roth et al.,1995). Anti-MOG antibody in combination with complement demyelinates mouse cerebellar slices as shown by loss of MBP-positive immunofluorescence and ELISA specific for MBP (Harrer et al.,2009). Theiler virus added to myelinated mouse spinal cord cultures caused marked demyelination (Shahar et al.,1986). In 2004, Birgbauer et al. developed a system of demyelinating rat cerebellar slices put into culture at P10 and demyelinated after 1 week using addition of lysolecithin to the culture medium for 15–17 h, with the subsequent return of myelin sheaths seen by immunofluorescence visible around 1 week later (Birgbauer et al.,2004). Measurement of CNPase activity in homogenized slices declined and was restored corresponding to demyelination and return of myelin sheaths, but it was not proven that this was remyelination of previously demyelinated fibers or myelination of new unmyelinated fibers. This is relevant as it is shown that Purkinje cells survive and produce new axons in these cultures (Dusart et al.,1997). Recently, however, we have shown definitively that the return of myelin sheaths truly represents remyelination as these sheaths are thinner (larger G ratios on EM) and have shorter internodal lengths, both of which are pathognomonic of remyelination (Zhang et al.,2011) (Fig. 3).

Figure 3. The myelin of remyelinated fibres in slice cultures is thinner and shorter than that found on fibers myelinated in development. (A)An electron micrograph showing myelinated fibers from a cerebellar slice culture. (B) An electron micrograph showing remyelinated fibers from a cerebellar slice culture. Myelin sheaths are thinner. Asterisks mark axons of a similar axon diameter, showing the reduction in myelin thickness in remyelinated fibers. (C) Quantification of G ratios (ratio of axon diameter to fiber diameter) of individual fibers in myelinated slices and remyelinated slices. Fibers in remyelinated slices have higher G-ratios per axon diameter, indicating thinner myelin sheaths. These regression lines are significantly different using the maximum likelihood ratio test, P <0.01. (D) Frequency distribution graphs of internode length (measured by distance between Caspr staining at paranodes) in myelinated and remyelinated fibres from cerebellar and spinal cord slice cultures. There is a left shift for remyelinated axons showing that these have shorter internodes. (P < 0.01 Kolmogorov–Smirnov Test). Scale bar = 1 μm. (Reproduced with permission from Zhang et al., Experimental Neurology,2011, 230, 138-148 © Elsevier).

Remyelination

Although remyelination (or at least return of myelin segments) was seen in slice cultures after demyelination of cerebellar slices as early as 1969 (Bornstein and Raine,1969), it is only more recently that the rate of remyelination has been altered by drug or genetic manipulations. Fingolimod, a drug now used in multiple sclerosis, has a positive effect on the return of myelin sheaths (measured by area of MBP-positive immunofluorescence after demyelination with lysolecithin) in mouse cerebellar slices (Miron et al.,2010). Inhibition of Lingo-1 has been shown to promote remyelination in newborn rat cerebellar cultures, after demyelination with lysolecithin, by MBP-positive immunofluorescence. Lingo-1 antagonism was also found to promote remyelination in vivo in rat spinal cord after LPC injection, suggesting that the behavior of slices reflects in vivo behavior (Mi et al.,2009). Similarly, addition of 9-cis retinoic acid enhances remyelination in this system, and its antagonists reduce remyelination, also seen in vivo, (Huang et al.,2011). This concordance of the effect on remyelination in the slice system ex vivo and in vivo is an obvious advantage of this model, and it has been confirmed using several other factors (Zhang et al.,2011).

Despite the demonstrated fidelity to the in vivo adult situation, a potential drawback of using slice cultures for study of remyelination is that the tissue is neonatal, and therefore may be more plastic than adult tissue. Adult slice culture has been attempted, with successful culture of adult rat hippocampal slices for 6 days using media supplemented with ATP and vitamins (Wilhelmi et al.,2002). Whole brain slices from adult rats aged up to postnatal day 40 also showed viable neurons after culture for 2 weeks by adding connexin-specific anti-sense oligodeoxynucleotides with the aim of reducing gap-junction mediated bystander cell death, but myelination was not assessed (Yoon et al.,2010). Success of remyelination in this slice system (as well as in vivo) clearly depends on the presence of viable neurons, though it is not yet clear what markers of neurodegeneration preclude successful remyelination.

QUANTIFICATION

The slice technique can be used to generate moderate throughput and high content screens for drugs/factors promoting or inhibiting myelination and remyelination in a context closely resembling that present in vivo. However, until recently, such screening has been stymied by a lack of a fast, accurate, reliable, objective, and automated system to quantify myelin sheaths.

Ideally, to quantify myelination, we would identify and measure the amount of a molecule only present in compact myelin, and absent from the myelinating oligodendrocyte cell body and processes. Furthermore, a molecule present in compact myelin of remyelinated sheaths but absent in normally myelinated sheaths would aid distinguishing the two processes. Unfortunately, neither such specific molecule has yet been identified, so alternative techniques have been used.

Initial approaches to quantify myelination relied on biochemical methods measuring myelin lipid and protein composites that altered in myelination, demyelination, and remyelination. Incorporation of [³⁵S] sulphate into sulfolipids in the presence of [³⁵S]SO₄ has been used both in vivo and in vitro/ex vivo (McKhann and Ho, 1967; Cardwell and Rome, 1988; Notterpek et al., 1993). High-performance thin-layer chromatography (HPTLC) densitometry was used to measure the lipid composition of organotypic cultures of mouse spinal cord, where the concentration of glycolipids (cerebrosides and sulfatides) is significantly decreased after demyelination, but not in a linear manner with the morphologic evaluation of demyelination (Roth and Bornstein, 1984). In cultures of Schwann cells and DRG neurons, myelination was tracked by following the uptake of a fluorescent ceramide analog and its appearance in the plasma membrane as fluorescent sphingolipid and galactocerebrosidase analog (Bilderback et al.,1997). However, this has not been carried out using oligodendroglial cells.

CNPase is an enzyme present in immature and mature oligodendrocytes and its activity has been used to assess myelination (Roth et al., 1983,1995; Roth and Bornstein, 1984; Vereyken et al.,2009). Alterations in CNPase expression levels appear earlier than light microscopic evidence of demyelination, as it is a marker of the degree of differentiation of oligodendrocytes present, rather than a marker of the presence of myelin sheaths. However, there was still a correlation of CNPase activity with the degree of myelination. Similarly, an approach using a β-galactosidase luminometric assay in a transgenic mouse expressing LacZ under the MBP promoter only measures MBP expression rather than the formation of myelin sheaths (Stankoff et al.,1996). Competitive inhibition ELISA was developed to measure the amount of MBP, found in both mature oligodendrocytes and the myelin sheath, in cerebellar explant cultures and found a 90% decrease in MBP on demyelination (Harrer et al.,2009; Nishimura et al.,1986). Myelin proteins, such as MBP, PLP/DM20, MAG, MOG, and CNPase, can also be measured by western blot from cultures (Brinkmann et al., 2008;Mi et al.,2009) as a surrogate marker of myelination.

These chemical analyses are generally simple, cheap and may be useful for screening, but are often only semi-quantitative, and may be insensitive in detecting minor changes in myelination and remyelination. In addition, as these proteins and lipids exist in mature oligodendrocyte cell bodies (whether myelinating or not), and in cell debris as well as in the myelin sheath, it is difficult to determine how meaningful these measurements are.

Immunohistochemistry has been used to try and improve myelination quantification. Use of antibodies against myelin markers such as MBP allows visualization of the myelin sheath, which can then be used to quantify the amount of myelin present. For example, the total MBP immunoreactive area divided by the scan area was used to describe myelination in tissue sections (McTigue et al., 1998; Miron et al.,2009) and slice culture (Miron et al.,2010; Mi et al.,2009). However, as MBP is present in not only the myelinating oligodendrocyte and myelin sheath, but also nonmyelinating oligodendrocytes and cell debris, overestimation of myelin sheath formation occurs. In addition, this method does not adjust for axon density; an increase in the MBP-positive immunoreactive area may simply be due to the presence of a greater number of axons. Recently, it was reported that all-trans retinoic acid increased MBP mRNA and protein expression in Schwann cell bodies in the peripheral nervous system, but reduced myelinated internodes (Latasa et al.,2010), indicating that MBP protein alone is not an ideal measure of myelination.

The need to consider axon density when quantifying myelination was addressed in the DRG-OPC co-culture system by relating the number of myelinating oligodendrocytes to the density of the underlying neurite network (Wang et al.,2007). However, this is time-consuming analysis, as although the neurite density can be calculated automatically by pixel counts, the percentage of MBP-positive myelinating oligodendrocytes must be counted manually.

Numbers of myelinated axons can also be visualised and counted by electron microscopy (Mi et al.,2009; Miron et al.,2010; Brinkmann et al., 2008; Vereyken et al.,2009). This is helpful in distinguishing remyelination from normally myelinated axons by measuring the thickness of myelin sheaths, but is expensive, slow and labor-intensive and is therefore not suitable for high-throughput analysis.

We have recently developed a method of quantifying myelination by measuring colocalization of MBP immunofluorescence with that of the high-molecular weight neurofilament subunit (NFH), markers of myelin and axons, respectively (Zhang et al.,2011). These measurements can be automated so that a confocal stack can take less than 10 s to analyze, making it a fast and objective method for quantifying myelination and remyelination; one with the possibility to be used effectively as a screening tool (Fig. 4A).

Figure 4. Quantification of aspects of myelination/remyelination. (A) Immunofluorescence for MBP (green) and NFH (red) on cerebellar slices shows co-localization where myelin sheaths myelinate axons (yellow). The color palette corresponding to overlap of signal is defined (within green line). All pixels of this color and intensity are shown, providing a “mask” outlining the myelin sheaths. The amount of myelin per axon (myelination index) is calculated as the mask area divided by the axon area (red). (B) Immunofluorescence for MBP (red), Caspr (green), and NFH (blue) can be used to categorize the stages of myelination in oligodendrocyte-DRG co-cultures. “Contacting”: oligodendrocyte processes contact axons but do not wrap them. “Extending”: oligodendrocyte processes are aligned with axons but not encircling them along the entire internode. “Wrapping”: oligodendrocyte processes encircle axons but Caspr is distributed throughout the internode. “Mature”: oligodendrocyte processes encircle axons but with Caspr restricted to paranodes. Scale bar = 10 μm. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

One question that is not addressed by this automated method of quantification is how effectively myelination has occurred in terms of thickness and compaction of the myelin sheath. Immunolabeling for axons surrounded by myelin shows only that oligodendrocytes have wrapped axons, but does not distinguish between initial wrapping and mature compact myelin. The gold standard for conclusively demonstrating that myelination is complete and compact at internodes flanked by formed paranodal loops is to use electron microscopy. However, a surrogate marker may be to observe the distribution of oligodendroglial protein neurofascin (NFC)-155 or its axonal binding partner Caspr, which mirrors it, by immunofluorescence. Early in myelination, these proteins are present at regions of axo-glial contact throughout the nascent internode. As myelination progresses, their distribution patterns coalesce into a spiral and later a compact band at each paranode (Pedraza et al.,2009). Using MBP and Caspr distribution as an indirect measure, each oligodendrocyte can be scored according to the extent to which it has myelinated underlying axons. One categorization scheme is to score oligodendrocytes as “contacting” axons, “extending” processes, “wrapping,” and “mature” (Huang et al.,2011) (Fig. 4B). Although this approach has the disadvantages of both being subjective and requiring manual analysis, it is still considerably more rapid and inexpensive than processing an equal number of samples using electron microscopy.

A thorough quantification of myelination could therefore include a large-scale automated quantification of the amount of myelin sheaths present, then scoring a subset of images for the stage to which myelination has proceeded using immunohistochemistry. Finally, if differences are observed, samples can then be analyzed in more detail using electron microscopy.

THE FUTURE

What is the future for these techniques? No doubt in vitro examination of pure oligodendrocyte cultures will continue, but we predict that there will be more use of co-cultures of CNS neurons and glia, and slice cultures, which preserve the three-dimensional structure and relationships of the brain. Automation is key to their further use in pharmaceutical testing/mass screening, and without this, these techniques are likely to remain in academic labs rather than in pharma. With the inevitable and important push towards fast development of translational therapies, faithful, rapid in vitro screening procedures will become essential tools.

Perhaps zebrafish are the ideal organism for study of myelination, as they are fast growing, easily manipulable and transparent as embryos. This readily allows invivo imaging of myelination in transgenic fish. Although studies in the zebrafish are already aiding our understanding of myelination (Kirby et al.,2006; Lyons et al.,2009), this may not all be directly translatable to mammals as the protein structure of the myelin sheath is different (CNS myelin in zebrafish contains myelin protein zero, which is only present in the PNS in mammals) and the zebrafish has a large Mauthner axon which has no equivalent in mammals. However, despite these caveats, zebrafish will be very powerful tools to provide fresh insights into myelination and, possibly, in the future, remyelination (Buckley et al.,2008).

Other ways to accelerate quantification of remyelination include designing transgenic mice so that myelin and neurofilament are labeled with different fluorescent markers, and that after demyelination, a genetic switch is used so that new myelin sheets are labeled with a third fluorescent marker. This could involve an inducible Cre recombinase to excise the first fluorescent protein sequence bringing the second into activity. Alternatively, a protein such as Kaede, which alters its color after exposure to ultraviolet light, could be used. It is known that proteins of compact myelin coupled to proteins such as GFP do not form normal compact myelin sheaths, and so small molecule labeling of such a protein, may be the ideal development.

Alternatively, myelination may be able to be assessed by nonfluorescent or non-immunohistochemical means, speeding up the analysis even further. One possibility is to use bioengineering techniques, by capitalizing on the fact that the mechanical properties of myelin are quite distinct from other cells, and may enable the amount to be assessed by changes of impedance, or infrared scattering (optical coherence tomography).

If we are to translate knowledge from study of myelination and remyelination into therapies for patients, then it may be useful to generate a system that uses human cells. Although, in the past, there have been slice cultures of human fetuses, this raises difficult ethical issues. Human embryonic stem cell technology may be useful to generate human neurons and human oligodendrocytes, which could be mixed in co-cultures. However, with the advent of induced pluripotent cell technology comes the ability to generate oligodendrocytes from patients with developmental or acquired diseases, and study them in culture, in many of the systems described above. The challenge currently is to generate a high-enough purity of such cells to be able to study them in culture, and then to determine whether their actions in culture represent their behavior in vivo.

Acknowledgements

HZ is supported by the China Scholarship Council. AAJ is supported by the Multiple Sclerosis Society of Canada and the UK Multiple Sclerosis Society. CFFC is supported by the Wellcome Trust, the MS Society UK, and the NMSS.