The development of new regenerative therapies for multiple sclerosis is hindered by the lack of potential targets for enhancing remyelination. The study of naturally regenerative processes such as the innate immune response represents a powerful approach for target discovery to solve this problem. By ‘mining’ these processes using transcriptional profiling we can identify candidate factors that can then be tested individually in clinically-relevant models of demyelination and remyelination. Here, therefore, we have examined a previously described in vivo model of the innate immune response in which zymosan-induced macrophage activation in the retina promotes myelin sheath formation by oligodendrocytes generated from transplanted precursor cells. While this model is not itself clinically relevant, it does provide a logical starting point for this study as factors that promote myelination must be present. Microarray analysis of zymosan-treated retinae identified several cytokines (CXCL13, endothelin 2, CCL20 and CXCL2) to be significantly upregulated. When tested in a cerebellar slice culture model, CXCL13 and endothelin 2 promoted myelination and endothelin 2 also promoted remyelination. In studies to identify the receptor responsible for this regenerative effect of endothelin 2, analysis of both remyelination following experimental demyelination and of different stages of multiple sclerosis lesions in human post-mortem tissue revealed high levels of endothelin receptor type B in oligodendrocyte lineage cells. Confirming a role for this receptor in remyelination, small molecule agonists and antagonists of endothelin receptor type B administered in slice cultures promoted and inhibited remyelination, respectively. Antagonists of endothelin receptor type B also inhibited remyelination of experimentally-generated demyelination in vivo. Our work therefore identifies endothelin 2 and the endothelin receptor type B as a regenerative pathway and suggests that endothelin receptor type B agonists represent a promising therapeutic approach to promote myelin regeneration.
The study of naturally occurring regenerative stimuli that promote remyelination provides a powerful strategy for discovering new targets to promote repair in multiple sclerosis. One such stimulus is the innate immune response. While CNS inflammation such as that seen in multiple sclerosis is usually thought to cause damage—a view supported by experimental models where the adaptive immune system is used to induce demyelination such as experimental autoimmune encephalomyelitis—the use of models in which the adaptive immune response is absent, such as toxin-induced demyelination, reveals a powerful regenerative role for the innate immune response (Kotter et al., 2001, 2005; Li et al., 2005; Chari et al., 2006; Glezer et al., 2006). As part of this response, a key role for macrophages is demonstrated by impaired remyelination following their depletion (Kotter et al., 2001). Two possible mechanisms by which macrophages enhance remyelination can be envisaged: the ability of macrophages to clear myelin debris, which is known to inhibit repair (Kotter et al., 2006; Ruckh et al., 2012) or secretion of regenerative factors such as cytokines. In support of the latter, we have shown that activation of the innate immune response by zymosan, a Toll-like receptor-2 ligand, increases the extent of myelination following transplantation of oligodendrocyte precursor cells into a CNS environment that lacks myelin—the nerve fibre layer of the retina (Setzu et al., 2006). Identification of these regenerative cytokines of the innate immune response therefore represents a logical next step in the search for therapeutic targets that enhance remyelination in multiple sclerosis.
While the experimental model of retinal inflammation and transplantation that we used above is clearly not a translational endpoint for studies to identify factors that enhance remyelination, it does provide an appropriate starting point for this next step. By performing a transcriptional analysis of the retinal innate immune response following zymosan injection (a response we know includes factors that enhance myelination by transplanted oligodendrocyte precursor cells) we can ‘mine’ this naturally regenerative process to assemble a list of candidate cytokines that can then be tested in other, progressively more complex models to identify those that do promote remyelination in areas of the CNS vulnerable to demyelination. We emphasize that the goal of such a study is target cytokine validation using clinically-relevant models of remyelination, not a detailed analysis of the (clinically-irrelevant) retinal myelination model that is used only to extract the initial list of candidate cytokines upregulated following induction of innate immunity. Using this strategy, we describe here the identification of endothelin 2 (ET-2, now known as EDN2) and the endothelin receptor type B (ET-B, now known as EDNRB) receptor as a novel therapeutic target and pathway for enhancing remyelination in multiple sclerosis.
Materials and methods
Retinal injection surgeries and analysis
All procedures were performed in compliance with UK Home Office regulations. Zymosan (12.5 µg/µl) or PBS was injected transvitreally into the retinae of adult female Fisher rats aged 9–14 weeks as previously described (Setzu et al., 2004, 2006). At various time points, retina samples were dissected out and cut in half. Half was fixed in 4% paraformaldehyde for histological analysis, and the other half preserved in RNAlater (Qiagen) for RNA extraction using a Qiagen RNeasy® Mini Kit.
Fixed retinae were blocked in 20% normal goat serum (Sigma) in 0.3% Triton PBS and then incubated overnight at 4°C with antibodies against CD68 (Serotec, 1:100) and GFAP (Dako, 1:100). Staining was visualized using appropriate secondary antibodies, and retinae were flat-mounted and examined using confocal microscopy.
Microarray and bioinformatics analysis
Microarray analysis was run on all samples using Illumina rat RefSeq chips containing 22 226 genes. To identify differentially-expressed genes, the statistical programming language R was used (http://cran.r-project/org/). Raw microarray data were log (base = 2) transformed and quantile normalized. The Tukey Box plot, covariance-based principal component analysis scatter plot, and correlation-based heat map were used to assure data quality. System noise was defined to equal the normalized expression value at which the locally weighted scatter plot smoothing fit of the observed coefficient of variation per gene by the observed mean expression per gene becomes non-linear. Genes not having at least one normalized expression value across all samples greater than system noise were discarded as ‘non-informative’. Genes not discarded were subject to the two-factor ANOVA (treatment × time). Genes found to have a Benjamini Hochberg false discovery rate multiple comparison corrected P-value < 0.05 were deemed ‘informative’ and subject to the Tukey Honestly Significant Difference post hoc test to identify differentially-expressed genes (uncorrected P-value < 0.05) at 3 days, 7 days and 14 days post-injection.
Oligodendrocyte precursor cell assays
Oligodendrocyte precursor cell cultures were prepared from neonatal Sprague Dawley rats as described previously (McCarthy and de Vellis, 1980; Laursen et al., 2009). Cells were plated with tested factors (ET-2 from Sigma, remainder from R&D Systems). To assess proliferation, cells were stained 48 h later for bromodeoxyuridene (BrdU Labeling and Detection Kit I, Roche) and Hoechst (bisBenzamide, Sigma). To assess survival or differentiation, cells were plated in SATO medium with factors and 0.5% foetal calf serum and allowed to differentiate for 72 h. Apoptotic cells were visualized using the ApopTag® Apoptosis Detection Kit (Millipore). To examine differentiation, cells were stained 72 h following plating for myelin basic protein (AbD Serotec) and Hoechst. In all analyses, five ×20 field images from each well plated were taken using an Axioplan Fluorescence microscope (Carl Zeiss) with Openlab imaging software. Three separate cell isolations were used, with three separate chambers from each isolation plated for each factor and dose. To examine ERK signalling, cultures were grown on 35 mm plastic dishes for 2–3 days then starved for 4–5 h before stimulation with ET-2 (10 ng/ml, Sigma) or the ET-B selective agonist BQ3020 (100 ng/ml, Sigma). For the antagonist experiment, cells were pretreated with the ET-B selective antagonist BQ788 (100 ng/ml) for 15 min before stimulation with BQ3020 for 15 min. Following treatment, cells were harvested and lysates for western blot were prepared as previously described by Gadea et al. (2008). Antibodies were diluted in 4% bovine serum albumin (Sigma) in Tris-buffered saline with 0.1% Tween 20, phospho-p44/42 MAPK (Erk1/2) (1:2000) and p44/42 MAPK (Erk1/2) (1:1000) (both from Cell Signaling Technology).
Cerebellar slice cultures
Following decapitation, brains of post-natal Day 0–1 CD1 mice were dissected out into Eagle’s medium with Earle’s salts medium (MEM). Sagittal slices (350 µm) of the cerebellum and underlying hindbrain were cut using a McIlwain tissue chopper, and then placed on Millipore Millicell-CM™ organotypic culture inserts in medium containing MEM, Earle’s balanced salt solution, heat-inactivated horse serum, GlutaMAX™, Fungizone®, and penicillin–streptomycin (each from Invitrogen), and glucose (Sigma). Factors were added at the desired concentrations, and fresh medium supplemented with factors every 2 days. Demyelination was induced after 14 days in culture without added factors by the addition of 0.5% lysolecithin (Sigma) to the medium for 20 h, after which slices were transferred back to either control medium or medium containing factors of interest, and maintained for an additional 14 days.
At 12 days in vitro for myelination and 28 days in vitro for remyelination, slices were fixed and stained. Confocal z-stacks were acquired (12 slices) with a Leica SPE confocal microscope and images analysed using NIH ImageJ. Only slices with intact cytoarchitecture were chosen for analysis, and imaging was focused on the areas of the cerebellum that contain parallel tracts of axons. Per cent area stained for Caspr or NFH was calculated and the degree of myelination or remyelination quantified by a ratio of per cent area stained for Caspr to per cent area stained for NFH. Using NIH ImageJ, Caspr images were carefully thresholded to only include areas of discrete, punctate paranodal Caspr staining, and to exclude more diffuse, internodal Caspr staining. Three separate slice isolations were used, with 8–12 slices analysed from each isolation for each factor and dose.
Cerebellar slice culture immunohistochemistry
Slices were fixed with 4% paraformaldehyde, blocked with 3% heat-inactivated horse serum, 2% bovine serum albumin, 0.25% Triton™ X-100, and then incubated in primary antibodies overnight at 4°C. Primary antibodies used were: anti-neurofilament 200 kDa (Encor Biotech, 1:4000), anti-myelin basic protein (AbD Serotec, 1:250), anti-Caspr (AbCam, 1:500), and anti-sodium channel (Sigma, 1:750). Staining was visualized by incubation in appropriate secondary antibodies at room temperature.
Cerebellar slice cultures were fixed overnight in 2.5% glutaraldehyde in 0.2 M sodium cacodylate and post-fixed in osmium tetroxide. The tissue was then dehydrated through ascending alcohols before embedding in resin. Cured resin blocks were trimmed and 1 µm sections cut, stained with toluidine blue and examined by light microscopy. Selected areas were further trimmed and ultrathin sections cut with a diamond blade onto copper grids using an ultramicrotome. Copper grid mounted ultrathin sections were stained with uranyl acetate and lead citrate and examined by transmission electron microscopy.
Demyelinating lesions were induced bilaterally by stereotactic injections of 4 µl of 0.01% ethidium bromide into the caudal cerebellar peduncles using a Hamilton syringe (Woodruff and Franklin, 1999). Female 3-month-old Sprague Dawley rats (180–210 g) were used for expression experiments and female 6-month-old rats (230–270 g) for delivery experiments. Non-radioactive in situ hybridization for ET-B in combination with immunostaining with antibodies against GFAP (Dako, 1:1000), CD68 (Serotec, 1:400), Olig2 (Millipore, 1:500) and Nkx2.2 (DHSB, 1:100) was then performed as described previously (Sim et al., 2002).
For delivery experiments, animals were administered 200 µl of a 23 nmol solution of BQ788, BQ3020, or saline by daily intraperitoneal injections from 7 days post-lesion to 14 days post-lesion (the time during which oligodendrocyte precursor cell differentiation is taking place). Animals were euthanized at 21 days post-lesion by perfusion with 4% glutaraldehyde under deep anaesthesia. The brain was dissected out, post-fixed and then cut into 1 mm transverse sections. Samples were embedded in resin, and 1 µm sections cut using a Leica RM2065 microtome and collected on glass slides. Following staining with 1% toluidine blue, sections from all three experimental conditions were ranked by an observer blind to these conditions for their level of remyelination as judged by light microscopy.
Multiple sclerosis lesion analysis
Brain tissue was obtained through a UK prospective donor scheme with full ethical approval from the UK Multiple Sclerosis Tissue Bank (MREC/02/2/39). Multiple sclerosis diagnosis was confirmed by Dr F. Roncaroli (Imperial College London) and clinical history was provided by Dr R. Nicholas (Imperial College London). Snap-frozen tissue was sectioned at 10 μm and classified according to the International Classification of Neurological Disease using Luxol fast blue staining and Oil Red O. We analysed five control blocks, six acute active lesions, four chronic active lesions, 14 chronic inactive lesions and 19 remyelinated lesions. Sections were fixed in 4% paraformaldehyde for 1 h, and permeabilized in methanol for 10 min at −20°C. Following antigen retrieval and blocking endogenous peroxidase activity (EnVision™ blocking solution, Dako), rabbit anti-endothelin B receptor primary antibody (1:200, Abcam) was applied overnight in a humid chamber at 4°C. Anti-mouse peroxidase conjugated secondary antibody was applied for 2 h at room temperature and stains visualized by 3,3’diaminobenzidene chromagen and light microscopy (Olympus BX61); sections were counterstained with Hoechst. For co-localization stains, sections were blocked with 10% normal horse serum and 0.1% Triton™ X-100 for 1 h, antibodies applied overnight as before, and fluorescently conjugated secondary antibodies applied for 2 h with Hoechst counterstain. Fluorescent images were acquired by confocal microscopy (Leica SPE). Immunopositive cells were counted per lesion and counts were multiplied accordingly to determine density of immunopositive cells/mm2 and statistical comparisons were done using Mann–Whitney test where P < 0.05 was considered to be statistically significant.
CXCL13, endothelin 2, CCL20 and CXCL2 are upregulated in an inflammatory environment supporting myelination
Our first goal was to assemble a list of candidate factors within the innate immune response that could be used for further studies on clinically relevant models to identify those that represent promising targets for enhancing regeneration. Thus, we used our previously described model of zymosan injection into the retina (Setzu et al., 2006). We performed a microarray analysis of gene expression at four time points (0, 3, 7 and 14 days post-injection) using the same zymosan injection protocol as in our previous work and using PBS injected retina at each time point as controls and Illumina Rat RefSeq chips for hybridization. An inbred rat strain (Fisher) was used to reduce variations due to genetic background, and three rats were used to provide three biological replicates at each time point and condition. Histological analysis (Fig. 1) confirmed changes in the retina consistent with an activation of the innate immune system. An increase in CD68-positive macrophages/activated microglia was seen with the greatest effect at 3 days post-injection. Increased GFAP expression, indicating a reactive response in retinal astrocytes (Müller cells), was also seen at all post-injection time points and was greatest at 14 days post-injection. Following microarray data normalization, we first excluded genes with an expression level below the noise level (15 086 of 22 226 genes), defined as the point at which the relationship between mean values and the coefficient of variation becomes non-linear. Then we used a two-factor ANOVA with a false discovery rate set at 5% to identify significant differences in gene responses. This analysis identified 44 upregulated genes at 3 days, eight upregulated genes at 7 days and 69 upregulated genes at 14 days (Supplementary Table 1). Within these differentially expressed genes, however, only four growth factors or cytokines were present: CXCL13, ET-2, CCL20 and CXCL2 (Supplementary Table 1). At 3 days, CCL20 was found to be upregulated 7.4-fold, CXCL2 4.8-fold, and ET-2 2.5-fold. None of these four genes were found to be significantly upregulated at 7 days. At 14 days, however, CXCL13 was found to be upregulated 16.1-fold, CCL20 13.4-fold, CXCL2 8.6-fold and ET-2 2.4-fold. The upregulation of CXCL13 and ET-2 following zymosan injection was validated using quantitative real-time PCR (Supplementary Fig. 1).
CXCL13 and endothelin 2 promote myelination
Next, we determined which of these candidate cytokines might provide potential targets for remyelination therapies by narrowing the list to include only those that enhanced myelination by endogenous oligodendrocyte precursor cells in normally-myelinated regions of the CNS. To do so, we used the simplest available assay in which oligodendrocyte precursor cells generate myelin-forming oligodendrocytes in intact CNS tissue—cerebellar slice cultures. Explant cultures from newborn mice were prepared by cutting sagittal sections of cerebellum and examined at 12 days in vitro (Fig. 2A). At this point myelination is not yet fully complete, allowing us to observe a potential acceleration by added factors. To quantify changes of myelination in response to added factors, we took advantage of the finding that the generation of compact myelin is associated with a change in the expression pattern of the axonal adhesion molecule Caspr, with exclusion from the internodal areas and its localization at the paranodal regions of the nodes of Ranvier (Eisenbach et al., 2009). This change in expression pattern was revealed in our slice cultures by immunofluorescent studies showing a concentration of voltage-dependant Na+ channels at the nodal axolemma and Caspr at the paranode (Fig. 2B), with electron microscopy directly confirming the presence of compact myelin at this time (Fig. 2C inset). As in our recent study examining the effects of a Wnt pathway inhibitor (Fancy et al., 2011), we used image analysis to quantify myelination in the slices by calculating the area of Caspr immunostaining corresponding to paranodes, using a thresholding technique to exclude the diffuse and weaker internodal Caspr immunostaining, and expressed this as a ratio of the area of axons. This unbiased approach allows accurate quantification of paranodal Caspr even in the presence of the variability of axon orientation and density observed in these cultures. This showed that CXCL13 and ET-2 promoted myelination significantly at 10 ng/ml, with CXCL13 also having a weak effect at 1 ng/ml (Fig. 2C and D). Neither factor was effective at the highest concentration (Fig. 2D). No effect was seen at any concentration of CCL20 or CXCL2 (Fig. 2D).
To determine whether this enhanced myelination represented an effect of CXCL13 and ET-2 that would increase the overall number of oligodendrocytes, we examined purified cultures of dissociated oligodendrocyte precursor cells and oligodendrocytes grown in the absence of axons and the presence of each of the four factors. We did not observe changes in proliferation, survival or differentiation [as measured by the expression of myelin basic protein (MBP)], each of which would be expected to increase oligodendrocyte numbers. Thus, quantification of proliferation by bromodeoxyuridene uptake showed that, apart from a very small reduction in proliferation seen at 1 ng/ml of CXCL13, none of the four factors altered oligodendrocyte precursor cell proliferation (Supplementary Fig. 2). Equally, we observed no changes in apoptosis in the cultures (Supplementary Fig. 3). Finally, there was no increase in the number of MBP+ cells generated within the cultures by CXCL13 or ET-2, although small effects were seen with the two factors that had no effect on myelination in the slices, CXCL2 and CCL20 (Supplementary Fig. 4).
Next, therefore, we tested an alternative explanation for enhanced myelination in the slices—an increased level of myelin formation by individual oligodendrocytes. We reasoned that myelin formation by an individual MBP+ oligodendrocyte could be measured by the formation of complex cellular processes and by the elaboration of membranous sheets of myelin. Thus, as in our recent study examining the effect of RXRγ signalling on oligodendrocyte differentiation (Huang et al., 2011), we performed a quantitative analysis of myelin sheet formation by scoring individual oligodendrocytes as having ‘simple’ (short, non-interdigitating processes), ‘complex’ (longer, interdigitating processes, but not membrane sheets), or ‘membrane’ morphology (processes containing MBP+ membrane sheets) (Fig. 2E). If a factor promotes myelin sheet formation, the number of cells with a simple morphology would decrease and the number of complex and membrane-forming cells would increase. Such changes were seen with ET-2 (at both 10 and 50 ng/ml) (Fig. 2E) and CXCL13 (at 10 ng/ml). The ratio of the most differentiated, membrane morphology cells increased from 0.107 ± 0.084 (mean ± SD) for controls to 0.480 ± 0.072 (ET-2, 10 ng/ml, P < 0.001), 0.232 ± 0.061 (ET-2, 50 ng/ml, P < 0.001), and 0.354 ± 0.088 (CXCL13, 10 ng/ml, P < 0.001). In contrast, there was no effect on oligodendrocyte morphology with the addition of CCL20 or CXCL2, showing that the stimulation of complex processes and myelin sheets in culture was a property only of CXCL13 and ET-2. We conclude from these experiments on cultured cells that the enhanced myelination seen in the slices reflects increased myelin formation by individual oligodendrocytes rather than any effect on proliferation, survival and differentiation.
Endothelin 2 enhances remyelination in explant cultures
Having narrowed the list of cytokines identified from the transcriptional profiling of the innate immune response to two factors that enhance myelination, we next examined whether these factors could also promote remyelination. To do this we modified the slice culture protocol above to allow myelination to take place for 14 days, after which there are extensive numbers of MBP+ internodes. Lysolecithin was then added for 20 h to demyelinate the slices, and then cultured for an additional 14 days during which remyelination occurs (Fig. 3A) (Miron et al., 2010; Huang et al., 2011; Zhang et al., 2011). Demyelination and remyelination (Fig. 3A and B) were visualized and quantified by using image analysis to measure the area of paranodal Caspr as above and as in our recent studies on Axin 2 (Fancy et al., 2011). In contrast with the experiments examining myelination in these slices, the addition of CXCL13 had no effect on remyelination (Fig. 3C and D). ET-2, however, enhanced remyelination at 10 ng/ml while 50 ng/ml had no effect (Fig. 3C and D). As in earlier studies of myelination, the formation of nodes of Ranvier and compact myelin in the cerebellar slices (here by remyelination) was confirmed with immunofluorescent and electron microscopy studies showing both demyelination following lysolecithin exposure and subsequent remyelination (Fig. 3E and F).
Endothelin receptor type B is required for remyelination
Having now narrowed the list of candidate cytokines to a single factor that promotes remyelination, ET-2, we next considered the question of which receptor is likely to be responsible. ET-2 is recognized by two G-protein coupled receptors termed endothelin receptors type A and B (ET-A and ET-B, respectively). In rodent cells of the oligodendrocyte lineage ET-B is more highly expressed than ET-A, as measured by immunofluorescence, western blotting and real-time-PCR of isolated cells (Gadea et al., 2009). ET-B is also highly expressed on human foetal oligodendrocyte precursor cells isolated by cell sorting and then subjected to transcriptional profiling (Sim et al., 2011). These isolated oligodendrocyte precursor cells are specifically those that were shown to have a high myelinogenic potential. ET-B signalling also makes a greater contribution than ET-A to endothelin receptor-induced phosphorylation of ERK (Gadea et al., 2009), the significance of which is highlighted by recent work showing that ERK2 is essential for timely myelination (Fyffe-Maricich et al., 2011). Together these suggest that ET-B signalling is likely to be responsible for the effects of ET-2 on oligodendrocytes that we observed above. To confirm this, we analysed the effects of a well-characterized ET-B-specific small molecule agonist (BQ3020) and antagonist (BQ788) (Ihara et al., 1992; Ishikawa et al., 1994) on ERK phosphorylation and on cell behaviour in the cell culture assays. As shown previously for endothelin 1 (ET-1) (Gadea et al., 2009) oligodendrocyte precursors treated with ET-2 showed rapid ERK phosphorylation (Supplementary Fig. 5). A similar effect was seen with the ET-B agonist BQ3020, and this could be blocked by the BQ788 antagonist (Supplementary Fig. 5). In the oligodendrocyte morphology assay, we found that the selective ET-B antagonist BQ788 inhibited myelin sheet formation at all concentrations tested (1–100 ng/ml), as revealed by a reduction in the number of cells with a differentiated, complex morphology and a corresponding increase in the fraction of those with a simple morphology from 0.212 ± 0.047 (control, mean ± SD) to 0.351 ± 0.028 (1 ng/ml, P < 0.001), 0.470 ± 0.041 (10 ng/ml, P < 0.001), 0.444 ± 0.033 (50 ng/ml, P < 0.001), and 0.433 ± 0.058 (100 ng/ml, P < 0.001) (Fig. 4A). In contrast, the selective ETB agonist BQ3020 increased the fraction of cells forming myelin sheets from 0.202 ± 0.042 (control, mean ± SD) to 0.335 ± 0.034 (50 ng/ml, P < 0.001) and 0.350 ± 0.032 (100 ng/ml, P < 0.001) (Fig. 4A). Activation of ET-B signalling therefore appears to mimic the effect of ET-2 on both ERK phosphorylation and on myelin formation, while the effect of the antagonist suggests that ET-B signalling in response to endogenous ligands contributes to the regulation of oligodendrocyte morphology during myelin formation.
Next, therefore, we determined whether ET-B was also expressed during remyelination in rodent models and, importantly for a study designed to identify potential targets for remyelination therapies, in human multiple sclerosis lesions. For the rodent studies we used a well-characterized model of remyelination, a focal lesion model in which demyelination is induced in the rat caudal cerebellar peduncle by injection of ethidium bromide (Fig. 4B) (Woodruff and Franklin, 1999; Sim et al., 2002). This in vivo model is advantageous for our study as there is no ongoing damage after the initial effects of the toxin and the tempo of the subsequent remyelination process is well defined; oligodendrocyte precursor cells repopulate the lesion at 5 days post-lesion, differentiate and initiate remyelination between 10–14 days post-lesion, and complete remyelination by 21–28 days post-lesion (Sim et al., 2002). First, we re-analysed our previously described microarray data set comparing gene expression profiles at 5, 14 and 28 days post-injection in the caudal cerebellar peduncle lesion. This showed a significant upregulation of ET-B between 5 and 14 days post-lesion (logFC 1.96, P = 0.01) corresponding to ∼4-fold increase. We confirmed that this increase represented expression on oligodendroglial cells by in situ hybridization studies of these lesions at 14 days post-lesion. These experiments showed that ET-B was expressed within lesions (Fig. 4B), and combined in situ hybridization and immunohistochemistry experiments showed that expression was present on Olig2+ oligodendroglial cells and Nkx2.2+/Olig2+ oligodendrocyte precursor cells, with macrophages/microglia also expressing the receptor (Fig. 4B). ET-B was also strongly expressed on cells around the lesion border where co-labelling with an antibody against GFAP indicated that many of these cells were astrocytes (Fig. 4B).
To establish the connection between our data in experimental animals and clinical disease in humans we examined four categories of multiple sclerosis lesion by immunohistochemistry: acute active, chronic active, chronic inactive and remyelinated, with both the rim and centre of the chronic active lesions examined separately (Fig. 5A). Increased numbers of ET-B expressing cells were seen in lesions (Fig. 5B), with this increase reaching significance for the rim of chronic active lesions and in remyelinated lesions (Fig. 5C)—both of which would be expected to contain oligodendrocyte precursor cells as they are areas of present or past regenerative activity. We confirmed directly that ET-B was expressed on oligodendrocyte precursor cells and differentiated oligodendrocytes in these lesions by demonstrating cells double labelled with Olig2 and NogoA, respectively (Fig. 5D). As in the experimental lesions, expression was also seen in macrophages/microglia and astrocytes, revealed by double labelling with CD68 and GFAP (Fig. 5D).
Having confirmed that both small molecule agonists or antagonists of ET-B promote or inhibit oligodendrocyte morphology changes that accompany myelination, and that ET-B is expressed on oligodendroglial cells during remyelination in rat and human, we asked whether agonists or antagonists of ET-B affected remyelination. In the ex vivo demyelinated cerebellar slice cultures analysed as above we found that blocking ET-B signalling with BQ788 at 10–100 ng/ml inhibited remyelination to levels below that seen in control slices, while activation of ET-B signalling with BQ3020 at 50–100 ng/ml enhanced remyelination (Fig. 6A and B). Interestingly, the level of remyelination seen with BQ3020 was greater than the maximal level observed with ET-2 (Fig. 3) suggesting that other ET-B ligands, such as ET-1, might also be able to contribute to remyelination through ET-B. Finally, we examined whether manipulating ET-B signalling could block or enhance CNS remyelination in vivo. Given the prior evidence from studies of horseradish peroxidase extravasation that the blood–brain barrier is defective in focal toxin lesions during remyelination (Felts and Smith, 1996), we used an intraperitoneal rather than an intracerebral route to administer the small molecule agonist and antagonist into the CNS. Following focal demyelination using the protocol described above, we injected saline, the ET-B agonist or ET-B antagonist intraperitoneally each day at 7–14 days post-lesion—the time of oligodendrocyte differentiation and initiation of remyelination. Animals were euthanized at 21 days post-lesion and semi-thin resin sections and ultrastructural analyses used to assess remyelination. We found that blocking ET-B signalling with the antagonist significantly impaired remyelination (Fig. 7A and B), although no effect of the agonist was observed in this model.
Here we show that ET-2 is a cytokine upregulated following activation of the innate immune response that enhances myelination and remyelination by oligodendrocytes. The ET-B mediates these regenerative effects. This receptor is expressed on oligodendrocyte precursor cells during remyelination in experimental rodent models and also on oligodendrocyte precursor cells within multiple sclerosis lesions, and agonists and antagonists of the receptor promote or inhibit myelination and remyelination in explant cultures. Moreover, ET-B activity is necessary for remyelination in vivo, as antagonists inhibit the normal regenerative response following toxin-induced demyelination in the adult rat. Together these results support the logic of our strategy of profiling the innate immune response to identify new targets for remyelination therapies, and identify endothelin signalling as a regenerative pathway in the CNS that leads to remyelination.
ET-2 belongs to the endothelin family of three 21 amino acid peptides: ET-1, -2 and -3 (Yanagisawa et al., 1988; Kawanabe and Nauli, 2011). ET-1 is widely expressed and implicated in a number of pathological processes including vasoconstriction, inflammation and tumour metastasis (Said et al., 2011). It is also expressed in the spinal cord of animals with experimental autoimmune encephalomyelitis, produced mainly by CD68-positive macrophages and activated microglia (Shin et al., 2001). ET-2 (the form we have identified in this study) is much less widely expressed and has not previously been reported either in the CNS or in association with inflammation, in contrast with ET-1 and -3, which are produced by both astrocytes and macrophages (Ehrenreich et al., 1990; MacCumber et al., 1990; Gadea et al., 2009). All these endothelins will bind to two endothelin receptors—ET-A and ET-B (Arai et al., 1990; Sakurai et al., 1990; Davenport, 2002). Both are G protein coupled receptors, but their association with different sets of G proteins results in their activation of separate downstream signalling pathways, for example, ET-A promotes vasoconstriction whereas ET-B can promote vasodilation (Aramori and Nakanishi, 1992; Rosano et al., 2009; Kawanabe and Nauli, 2011). Within the CNS, a role in pain sensation is well described (Khodorova et al., 2009), and here again the two receptors may have differing functions as the expression pattern of ET-A and ET-B in sensory nerves is distinct, being on small diameter axons and satellite glial cells, respectively (Pomonis et al., 2001).
Previous studies of endothelin signalling have suggested a possible indirect effect on regeneration, as ET-B is highly expressed on astrocytes around the border of the demyelinated lesion, as shown in this study, and both ET-1 (Gadea et al., 2008) and ET-B signalling (Koyama et al., 1999; Rogers et al., 2003) promote astrogliosis. These activated astrocytes may therefore release extracellular matrix and/or soluble factors that promote remyelination. Evidence for a direct effect of endothelin signalling on oligodendrocyte lineage cells came first from a developmental study showing that ET-1 secreted by astrocytes promotes oligodendrocyte precursor cell migration and inhibits the acquisition of differentiation markers such as MBP (Gadea et al., 2009). Experiments with receptor agonists and antagonists revealed a more significant role for ET-B than ET-A in migration (Gadea et al., 2009). Our study now shows that ET-B also has a role in remyelination, and increases myelin formation by cultured oligodendrocytes. In contrast, neither ET-1 (Gadea et al., 2009) or ET-2 appear to have a role in the regulation of proliferation or survival of oligodendroglial cells during remyelination. Together these studies highlight the potential importance of endothelin signalling and ET-B in an example of CNS regeneration—remyelination—by directly enhancing at least two of the steps required: the recruitment of precursors and the elaboration of myelin by differentiated cells. This conclusion as to the role of endothelin signalling is further supported by our experiment showing that inhibition of ET-B by a peripherally-administered small molecule antagonist impairs remyelination of focal lesions in vivo. The lack of any effect of the ET-B agonists in vivo remains to be explained and may reflect inappropriate dosing and/or an effect of ET-B signalling on other tissues such as blood vessels that prevent a regenerative response; however, it should be noted that remyelination in young adult rodents as used here is already extremely efficient and it may therefore be difficult to reveal any beneficial effect.
One discrepancy between our study using ET-2 and the previous work above using ET-1 is the effect on oligodendrocyte differentiation. We found no effect of ET-2 on the number of MBP+ cells while Gadea et al. (2009) found an inhibition of MBP expression with ET-1. Given that both endothelins bind to the same receptors, this was unexpected. One possible mechanism could be the presence of distinct co-receptors associating with each endothelin receptor that bind to one or the other endothelin. Another possibility is the triggering of distinct downstream pathways dependent on the level of ligand binding and receptor activation. An example of this is provided by another growth factor receptor, PDGFRα, where high and low levels of signalling result in the activation of PI3K (needed for oligodendrocyte precursor cell proliferation and migration) or PLCγ signalling (only needed for oligodendrocyte precursor cell migration), respectively (McKinnon et al., 2005). However, the observation from another study of ET-1 showing increased MBP in cultured oligodendrocytes (Jung et al., 2011) in apparent contradiction to the results of Gadea et al. (2009), suggests additional complexity that further experiments are required to unravel.
Our finding that ET-2 and ET-B signalling promote the changes in shape that lead to myelin membrane formation, but have no effect on the number of cells expressing myelin proteins such as MBP, is interesting in that it highlights the distinct regulation of two components of oligodendrocyte biology that are often considered as a single process—differentiation and myelin formation. From a cell biology perspective, however, linking regulation of shape and gene expression is not necessarily appropriate given the very different molecular mechanisms involved, and our work adds to other observations that emphasize this (Rosenberg et al., 2008). So, for example, myelination in rat optic nerve begins at the retinal end of the nerve whereas myelin protein expression is first seen in oligodendrocytes at the chiasmal end (Colello et al., 1995). Experiments in co-culture of neurons and oligodendrocytes show that NGF inhibits myelination by already differentiated MBP+ oligodendrocytes (Chan et al., 2004), and a similar result is seen with PDGF and FGF2 (Wang et al., 2007). Together with experiments showing that the extracellular matrix substrate merosin promotes myelin sheet formation by cells cultured without axons without any effect on the number of MBP+ cells (Buttery and ffrench-Constant, 1999), these results clearly show that the pathways controlling shape and gene expression can be regulated directly and independently in oligodendrocytes.
As well as identifying a novel signalling pathway that promotes remyelination, our present findings add significantly to our understanding of how inflammation might promote regeneration. Earlier studies of the CNS have identified regenerative properties of inflammation by the description of macrophage-derived oncomodulin as a factor that promotes retinal ganglion axon regeneration (Yin et al., 2006, 2009) and by the finding that mice lacking inflammatory cytokines IL1β or TNFα have delayed remyelination following demyelination caused by the administration of cuprizone (Arnett et al., 2001; Mason et al., 2001). Taken together with these studies, our results emphasize how the inflammatory response represents a rich source of potential novel regenerative factors for therapeutic exploration.
This work was supported by the United Kingdom Multiple Sclerosis Society (to C.ff.C., R.J.M.F.); National Multiple Sclerosis Society (to C.ff.C., R.J.M.F.); and the Wellcome Trust (to C.ff.C.) T.J.Y. was supported by the National Institutes of Health-University of Cambridge Graduate Partnerships Program and the International Biomedical Research Alliance.
Supplementary material is available at Brain online.
endothelin receptor type A
endothelin receptor type B
The authors would like to thank Andrew Jarjour, Lisbeth Laursen, David Coutts, and Anna Setzu for technical assistance.