The formation of a glial scar does not prohibit remyelination in an animal model of multiple sclerosis

Abstract The role of astrocytes in the pathophysiology of multiple sclerosis (MS) is discussed controversially. Especially the formation of the glial scar is often believed to act as a barrier for remyelination. At the same time, astrocytes are known to produce factors that influence oligodendrocyte precursor cell (OPC) survival. To explore these mechanisms, we investigated the astrocytic reaction in an animal model induced by immunization with myelin oligodendrocyte glycoprotein (MOG) in Dark Agouti (DA) rats, which mimics most of the histological features of MS. We correlated the astroglial reaction by immunohistochemistry (IHC) for glial fibrillary acidic protein (GFAP) to the remyelination capacity by in situ hybridization for mRNA of proteolipid protein (PLP), indicative of OPCs, over the full course of the disease. PLP mRNA peaked in early remyelinating lesions while the amount of GFAP positive astrocytes was highest in remyelinated lesions. In shadow plaques, we found at the same time all features of a glial scar and numbers of OPCs and mature oligodendrocytes, which were nearly equal to that in unaffected white matter areas. To assess the plaque environment, we furthermore quantitatively analyzed factors expressed by astrocytes previously suggested to influence remyelination. From our data, we conclude that remyelination occurs despite an abundant glial reaction in this animal model. The different patterns of astrocytic factors and the occurrence of different astrocytic phenotypes during lesion evolution furthermore indicate a finely regulated, balanced astrocytic involvement leading to successful repair.

tion in an animal model induced by immunization with myelin oligodendrocyte glycoprotein (MOG) in Dark Agouti (DA) rats, which mimics most of the histological features of MS. We correlated the astroglial reaction by immunohistochemistry (IHC) for glial fibrillary acidic protein (GFAP) to the remyelination capacity by in situ hybridization for mRNA of proteolipid protein (PLP), indicative of OPCs, over the full course of the disease. PLP mRNA peaked in early remyelinating lesions while the amount of GFAP positive astrocytes was highest in remyelinated lesions. In shadow plaques, we found at the same time all features of a glial scar and numbers of OPCs and mature oligodendrocytes, which were nearly equal to that in unaffected white matter areas. To assess the plaque environment, we furthermore quantitatively analyzed factors expressed by astrocytes previously suggested to influence remyelination. From our data, we conclude that remyelination occurs despite an abundant glial reaction in this animal model. The different patterns of astrocytic factors and the occurrence of different astrocytic phenotypes during lesion evolution furthermore indicate a finely regulated, balanced astrocytic involvement leading to successful repair.

K E Y W O R D S
astrocytes, glial scar, multiple sclerosis, oligodendrocyte precursor cells, remyelination, shadow plaque

| INTRODUCTION
Astrocytes are the most abundant cell type in the mammalian brain, outnumbering neurons by over fivefold (Sofroniew & Vinters, 2010).
In the healthy central nervous system (CNS), astrocytes play important roles in supporting synaptic transmission and providing functional support of neurons by maintaining local ion and pH homeostasis, storing CNS glycogen, clearing neuronal waste, and providing maintenance and modulation of the blood-brain barrier (Nair, Frederick, & Miller, 2008). Astrocytes react to all sorts of insults to the CNS in a process referred to as reactive astrogliosis, which is often seen as pathological hallmark of diseased CNS tissue. These changes range from rather mild reversible alterations with cellular hypertrophy without overlap of astrocyte processes to long-lasting scar formation. The glial scar is associated with a rearrangement of tissue structure, astrocyte proliferation, and a pronounced overlap of astrocyte processes, which result in a disruption of individual astrocyte domains (Sofroniew & Vinters, 2010). Pathologically the hallmark of astrogliosis and activated astrocytes is an upregulation of glial fibrillary acidic protein (GFAP), which belongs to the family of the intermediate filament proteins and is commonly used as a marker for astrocytes in histology.
Mild to moderate astrogliosis is believed to protect CNS cells and tissue by uptake of potentially excitotoxic glutamate, protects from oxidative stress, facilitates the blood-brain barrier repair and limits the spread of inflammatory cells or infectious agents from areas of damage or disease into healthy CNS tissue (Iglesias, Morales, & Barreto, 2017;Ludwin, Rao, Moore, & Antel, 2016;Nair et al., 2008;Sofroniew & Vinters, 2010). In contrast, the formation of a compact glial scar is believed to act as an inhibitor of regeneration because of its mechanical rigidity, especially in the scenario of multiple sclerosis (MS) (Brambilla et al., 2014;Ghatak, 1992). This has even led to experimental approaches which aimed at eliminating reactive astrocytes or prevent them from becoming fully reactive to foster regeneration (Correale & Farez, 2015). Contrariwise astrocytes were shown to produce several factors, which facilitate repair processes upon activation. The most common model for studying MS is experimental autoimmune encephalomyelitis (EAE) induced in rodents by active immunization with myelin oligodendrocyte glycoprotein (MOG) because it mimics many of the pathophysiological mechanisms and histopathological features of this disease. Histologically it allows to study the full course of lesion evolution from active demyelination up to successful remyelination, the shadow plaque (SP) (Gold, Hartung, & Toyka, 2000;Weissert et al., 1998). The aim of our study was to investigate the astrocytic reaction at different stages of lesion evolution and the correlation between the extent of astrogliosis and remyelination in MOG EAE. To gain more information about the role of astrocytes in remyelination, we furthermore investigated the distribution of a selected set of astrocyte-derived factors known for their involvement during lesion evolution. We hypothesized that a complex arrangement of these factors is needed at specific time points to result in successful remyelination. Furthermore, we followed the two astrocytic phenotypes A1 and A2 during lesion evolution where the A1 phenotype has been suggested to prohibit remyelination, whereas A2 appears to be the remyelination promoting phenotype (Liddelow et al., 2017). We discuss the expression pattern seen in our experimental sample of successful remyelination as opposed to the available data in human MS brain specimen, where often remyelination is not successfully accomplished. Understanding the role of astrocytes in remyelination could lead to a better knowledge of repair mechanisms and how they could be influenced.

| Used material, selection of lesions, and definition of lesional staging
The study was performed on archival material from previous EAE experiments in female DA/OlaHsd rats 8-10 weeks actively immunized with recombinant rat MOG (rrMOG), corresponding to the N-terminal sequence of rat MOG (aa1-125) collected within the archives of the neuroimmunology research laboratory between 1998 until present (Hochmeister et al., 2012;Seifert, Bauer, Weissert, Fazekas, & Storch, 2007;Storch et al., 2006). Detailed information on clinical course of the disease is given in the respective references Weissert et al., 1998Weissert et al., , 2003. For the purpose of our investigation, we selected material from animals with clinical symptoms and large confluent lesions which led to a total of 45 animals.
Lesions were classified as active lesion (A), inactive to early remyelinating lesion (IA/ER), early to late remyelinating lesion (ER/LR), late remyelinating lesion to SP (LR/SP) and SP following previously suggested criteria (Kornek et al., 2000;Lucchinetti et al., 1999). In short, actively demyelinating lesions are defined by the presence of macrophages containing luxol fast blue (LFB) positive myelin degradation products, which are immunoreactive for myelin proteins. In early active lesions, these degradation products are immunoreactive for all myelin proteins including MOG, whereas in late active lesions the degradation products are immunoreactive for the major myelin proteins, myelin binding protein (MBP) and proteolipid protein (PLP) but not for MOG. We validated myelin degradation products within macrophages with immunofluorescent (IF) double staining against CD68 (ED1, a marker for macrophages) and MOG. In inactive demyelinated lesions, macrophages may be loaded with PAS-positive granular or sudanophilic lipids but do not contain myelin degradation products (Lucchinetti, Brück, Rodriguez, & Lassmann, 1996). Early remyelinating lesions are similar to inactive demyelinated plaques but contain myelin fibers positive for certain myelin proteins. Late remyelinating lesions (classical SPs) show a reduced staining intensity (myelin pallor) in conventional myelin staining (LFB) compared to the periplaque white matter due to thin, newly formed myelin sheaths. Areas without any histopathological changes, that is normal appearing white matter (NAWM), served as control regions (Zeis, Graumann, Reynolds, & Schaeren-Wiemers, 2008).

| Tissue processing, immunohistochemistry and immunofluorescence
Serial paraffin sections from 1.5 to 2 μm were prepared (for in situ hybridization under RNase-free conditions) and dewaxed in Xylene (Roth, Karlsruhe, Germany). Slides were stained for hematoxylin & eosin (H & E) and LFB to obtain a first overview on tissue inflammation and demyelination.
For immunohistochemistry (IHC), slides were dewaxed, rehydrated, and incubated in a methanol solution containing H 2 O 2 (Merck, Darmstadt, Germany) to be able to block the endogenous peroxidase (Bauer et al., 2002). Antigen retrieval was performed using EDTA (Sigma, St. Louis, Missouri, USA) at pH 5.2 in a steamer for 1 hr. After a cooling step, slides were washed and blocked for 20 min with 10% fetal calf serum (FCS; Sigma) in DAKO REAL Antibody diluent (Dako; Santa Clara, CA). The blocking solution was discarded and the primary antibody was applied at a dilution of 1:100 in FCS/DAKO and incubated over night at 4 C. After three washing steps with PBS, the secondary biotinylated antibody (GE Healthcare, Buckinghamshire, Great Britain) was applied at a dilution of 1:200 in FCS/DAKO and incubated at room temperature for 1 hr. Coated with peroxidase labeled avidin (Sigma) diluted 1:100 in FCS/DAKO slides were incubated at room temperature for 1 hr. Visualization of the primary antibody was then performed with 3,3 0 -diaminobenzidine (DAB; Sigma) and counterstaining with hematoxylin (Gatt-Koller, Absam, Austria). Slides were dehydrated in ascending ethanol concentration and covered with Shandon Consul-Mount (Thermo Scientific, Waltham, Massachusetts, USA) and a cover slip.
For IHC double staining, pretreatment procedures were performed as described earlier. GFAP and one astrocytic factor, respectively, were incubated simultaneously in FCS/DAKO at 4 C overnight on selected lesions. For the detection of primary antibodies, a polymer-based detection system (ImmPress; Vector laboratories, Burlingame, CA) was used according to the manufacturer's instructions.
Visualization of anti-mouse ImmPress reagent was performed with DAB (Sigma) and visualization of anti-rabbit ImmPress or anti-goat ImmPress reagent, respectively, were obtained with VIP Peroxidase substrate kit (Vector Laboratories). Afterward, slides were dehydrated and covered with Shandon Consul-Mount (Thermo Scientific) and a coverslip. Further information about all used antibodies can be found in the Supporting Information Table S1.  with 10% FCS for 15 min. Anti-digoxigenin antibodies conjugated with alkaline phosphatase (Roche) were applied at a dilution of 1:500 in blocking reagent with 10% FCS for 1 hr. The sections were developed with NBT/BCIP (Roche) at 4 C under microscopic controls.

| In situ hybridization/IHC double staining
For the immunohistochemical double staining, mouse PLP primary antibodies were applied at a dilution of 1:1000 in 10% FCS/TBS over night at 4 C. After some washing steps, the biotinylated secondary mouse antibody (GE Healthcare, Buckinghamshire, Great Britain) was applied at a dilution of 1:200 in FCS/TBS 1 hr at room temperature.
Then incubation with avidin-alkaline phosphatase complex (Roche) was performed for 1 hr at room temperature. The slides were developed using a fast red solution (Sigma). Counterstaining was performed with hematoxylin and slides were coverslipped as described earlier.

| Histopathology and quantitative evaluation
The number and types of lesions and the number of rats which were available for GFAP-IHC, PLP-ISH, and NOGO-IHC are given in Table 1. Analysis was performed by a single investigator blinded for lesion types (MTH). The selected lesions were quantitatively assessed on a Zeiss Axioplan2 imaging light microscope under a 400× magnification. All specifically stained cells for the respective antigen were quantitatively assessed in one full square grid and results converted into corresponding cells per square millimeter. Table 1, the number of assessed lesions in GFAP-IHC and PLP-ISH varies between the staining types due to tissue loss or unspecific results during the staining process. These staining problems appear among others due to the polymerization of GFAP during astrocytic reaction, masking the epitopes (Ludwin et al., 2016). For NOGO-IHC, we used a selected set of lesions. We investigated the lesion evolution in brain (cerebellum) and spinal cord specimens. For quantitative evaluation and statistical analysis, we only used spinal cord lesions to ensure comparability; however, there is a close histopathological resemblance of the lesions in the spinal cord and in the brain.

As shown in
We chose GFAP as marker for astrocytes because of its best countability. For quality control, we stained two additional markers for astrocytes, Nestin and Vimentin. In Supporting Information Figure S1a-i, the three different markers are shown on the same lesions side-by-side for comparison. We also performed double staining of GLT1 (Excitatory Amino Acid Transporter 2 EAAT2, rodent analog GLT1) and GFAP on ER/LR lesions (Supporting Information Figure   S1j). As postulated by Yang et al., 2011, GLT1 is primarily distributed at the membrane of astroglial processes thereby making it difficult to clearly identify individual astrocytes by their immunostaining (Yang et al., 2011). As GFAP is an intermediate filament the whole cell body is stained quite well, which makes it more practicable during quantification. In total, GFAP immunoreactivity (GFAP-IHC) was assessed in 540 lesions, OPC density (PLP-ISH) in 450 lesions, and occurrence of mature Oligodendrocytes (NOGO-IHC) in 120 lesions by one blinded investigator (MTH).
As the lesion characterization serves as basis of all presented research, we validated our characterization with IF staining against macrophages (ED1) and myelin components (MOG) as most of the characterization depends on the presence or absence of myelin degradation products within macrophages in lesions (Supporting Information Figure S1k,l).

| Evaluation of astrocytic phenotypes A1 and A2
To follow the two different astrocytic phenotypes A1 and A2 over the full course of lesion evolution, we chose the markers described by Liddelow et al., 2017; C3d complement for detecting A1 and S100 calcium-binding protein A10 (S100A10) for A2 (Liddelow et al., 2017).
We documented the results via selected microscopy pictures. In a correlation analysis, we correlated OPC density with the respective factor occurrence during lesion evolution (A -SP). For this purpose, we selected the same lesions for OPC as selected for all analyzed factors and used the Spearman-Rho nonparametric correlation with p < .01 considered as significant (two-tailed).

| Histopathological changes in rat brains and spinal cords
White matter lesions in the brains (usually located in the cerebellar white matter and occasionally in the white matter of the corpus callosum) and-more often-in the spinal cords of DA rats as typically caused by the MOG EAE model were identified and analyzed with respect to lesion types as described earlier (Table 1).
Adjacent slides of each lesion were subjected to IHC for GFAP to detect the astrocytic reaction and ISH for PLP mRNA to monitor   ing remyelination process. We arranged the factors according to their peak occurrence during lesion evolution and discuss the results systematically over the full course of lesion evolution. Figure 5 gives an overview of the appropriate lesions fitting to our statistical results.
During active demyelination, we found Sema3a, Sema3F, HABP2 and BMP-2 to be highly up-regulated (Figure 4a-d; Figure 5a-   Astrogliosis as a marker of diseased tissue increases during lesion evolution and reaches its peak in SP representing all features known for a glial scar (a). In NAWM, only a few astrocytes are GFAP positive and a significant upregulation of GFAP expressing astrocytes is detectable in SP with almost all astrocytes expressing GFAP. All groups differ significantly from each other (p < .01). In total, 540 lesions were quantified for GFAP (out of 45 rats). The observation of OPCs during lesion evolution reveals in a peak of PLP mRNA expression in IA/ER (b). In A hardly any OPCs expressing PLP mRNA are detectable. PLP mRNA expression in fully remyelinated SP shows close resemblance to the expression in NAWM. ER/LR and LR/SP results are slightly over the range of detected OPCs in NAWM and SP. Only A and IA/ER lesions differ significantly from the other groups (p < .01). In total, 450 lesions were quantified for PLP (out of 45 rats). As expected the distribution of NOGO-positive cells (mature oligodendrocytes) follow the course of OPCs with a slight delay. During A mature oligodendrocytes are comparably low. There is an increase in mature oligodendrocytes detectable during early remyelination indicating a successful differentiation of OPCs. Mature oligodendrocytes are also detectable during late remyelination and their occurrence in SP is comparable with their appearance in NAWM. All groups differ significantly from each other (p < .01) with the exception of NAWM to SP (p = .541) and IA/ER to ER/LR (p = .086). In total, 120 lesions were quantified for NOGO (out of 15 rats) FIGURE 4 Percentage scale of GFAP positive astrocytes and astrocyte-derived factors on selected spinal cord lesions. In each lesion type the mean of all astrocytes represents the 100% scale and the corresponding astrocytes expressing factors involved in OPC regulation are given in the appropriate ratios (for each factor n = 20 selected lesions per type). Error bars represent 95% CI. The guidance molecule Sema3A (a) peaked in A with more than 60% of active astrocytes expressing this factor and additionally showed a high occurrence in LR/SP with approximately 55%. In A, the expression of Sema3F (b) is higher than in IA/ER and peaks again in ER/LR. HABP2 (c) is highly present during A with more than 80% of all astrocytes expressing this factor. It decreases again to 5-30% during remyelination. BMP-2 (d) is comparably high during A decreases in IA/ER and increases again during later remyelination steps. CNTF expression (e) increases during lesion evolution with a peak of almost 70% CNTF positive astrocytes in IA/ER and decreases again to approximately 5% in SP. BDNF (f ) expression increases from NAWM to IA/ER, decreases in ER/LR, increases again in LR/SP and reaches its minimum in SP. CXCL-12 (g) increases during lesion evolution, peaks in ER/LR with 80% of active astrocytes expressing this factor and decreases again from LR/SP to SP. Around 50% of all astrocytes are FN1 positive during A and late remyelination steps (h). Tn-C is upregulated during A and ER/LR (i). The expression of IGF (j) rises from NAWM to LR/SP with a slight increase in A and falls rapidly from LR/SP to SP. FGF-2 (k) showed only a slight increase from NAWM to A and decreases from LR/SP to SP. IL-6 (l) shows a more uniform course with a slight increase of IL-6 expression during remyelination process with more than 50% of all active astrocytes expressing IL-6

| DISCUSSION
Remyelination of the CNS after a demyelinating event proceeds in distinct stages. OPCs are recruited into demyelinated areas and differentiate into myelinating oligodendrocytes to form new myelin sheaths (Horner & Gage, 2000;Kotter, Stadelmann, & Hartung, 2011). PLP is one of the major proteins in compact myelin and its expression occurs  (Hagemeier, Brück, & Kuhlmann, 2012;Itoh et al., 2018;Nair et al., 2008;Sofroniew & Vinters, 2010), but especially the formation of a dense glial scar is often believed to act as an impenetrable mechanical barrier for any remyelination efforts.
In this experimental study, we show that in our sample of MOG EAE in DA rats, the astrocytic reaction increases over the time course  (Nair et al., 2008). The number of OPCs in this lesion area however equaled the range of OPCs found in the NAWM, indicating that OPCs, on the contrary, must be able to migrate and proliferate very well even in this environment. This is furthermore supported by our investigation of mature oligodendrocytes in the same sample-an equal number of oligodendrocytes was found in NAWM and SP.
By tracing OPCs via ISH for PLP mRNA as well as mature oligodendrocytes over the full lesion course, we found that after a nearly complete loss of OPCs during active demyelination they are recruited back again once the lesion becomes inactive. As anticipated, the highest amounts of PLP mRNA, representing increased remyelinating activity of OPCs, were found in IA/ER lesions. Upon successfully completed remyelination, PLP mRNA decreased again. The remarkable decrease in PLP mRNA expressing OPCs from IA/ER to later remyelination stages is explainable by the maturation of OPCs to myelinating oligodendrocytes. Accordingly, the highest number of mature oligodendrocytes was detectable in ER/LR. The SP exhibited an OPC and mature oligodendrocyte number as seen in the NAWM despite showing features of a dense glial scarring. We concluded, therefore, that extensive glial reaction and even scar formation does not per se hinder migration and survival of OPCs but, on the contrary, might even allow and support remyelination in the presence of the right mediators facilitating repair at the right lesion stage.
To test this hypothesis, we investigated the expression of several astrocyte-derived factors with assumed influence on remyelination over the full course of lesion evolution to gain more insight into the plaque environment at every lesion stage. Based on the available literature, we concentrated on a subset of exemplary astrocytic factors and quantitatively assessed the percentage of astrocytes in the lesions expressing the respective factor.
The first step for a successful repair is the appropriate recruitment of OPCs into the lesion itself. There are several factors known to be necessary for OPC guidance. We found factors with different properties and functions at this stage. Sema3A and Sema3F are on the one hand known in the literature as guidance molecules for axons and OPCs in the developing nervous system and are also upregulated in MS brains in glial cells and neurons in the vicinity of active demyelinating lesions. Especially the expression of Sema3F is believed to serve as a recruitment signal for OPCs into the lesion (Williams, de Wit, & Ghosh, 2010). In contrast, both Semaphorines are also discussed as reversible inhibitors of OPC differentiation (Syed et al., 2011). In our sample, we detect high levels of both Sema3A and Sema3F expression in all lesion types with a peak of Sema3A in A and LR and of Sema3F in ER/LR, obviously representing sequential recruitment signals on OPCs during remyelination, at the same time maybe preventing too early differentiation of OPCs. We hypothesize that besides a guidance function also the inhibition of OPC differentiation is necessary at this lesion stage. As OPCs have to be recruited first, a too early differentiation might be detrimental for migration to the particular lesion.
The second factor of importance at this lesion stage synthesized by astrocytes and analyzed in our work was hyaluronan (Back et al., 2005). Hyaluronan deposits accumulate in demyelinated lesions in MS patients and in mice with EAE and are discussed to inhibit OPC maturation (Sloane et al., 2010). In our sample, we found hyaluronan to be present in 80% of astrocytes during active demyelination followed by a decrease during remyelination. This finding is in line with the assumption that the blocking effect of hyaluronan on OPCs seems to be reversible and controlled by enzymatic degradation (Sloane et al., 2010) and supports our hypothesis that blocking of maturation is necessary during guidance and migration steps.
Another factor described during the stage of OPC guidance is BMP-2. It is thought to promote astrocyte reactivity and stimulate glial scar formation. BMPs in general are discussed to decrease proliferation of precursors, increase apoptosis, and increase expression of GFAP. Increase in BMP-2 is discussed to be an early signal stimulating glial scar formation after demyelinating injuries (Fuller et al., 2007). It is furthermore described to inhibit the production of oligodendrocytes in the healthy CNS. As BMPs are expressed in demyelinating lesions in human brains, it is even described to be a potential therapeutic target to enhance repair (Sabo, Aumann, Merlo, Kilpatrick, & Cate, 2011).
In our sample, we found BMP-2 mainly during active demyelination and late remyelination, but it was downregulated during early remyelination. The rather negative view of BMP-2 regarding its impact on remyelination in the literature most likely correlates with the-most often-negative connotation of the glial scar. When describing BMP-2 as a promotor of astrogliosis, it is often described as inhibitor of remyelination. We suggest here a rather supportive role of the glial scar than a detrimental one, which may also changes the view of BMP-2.
The next mandatory step for successful remyelination is the migration of OPCs into the lesion area and the associated start of repair. We therefore followed the two factors CNTF and BDNF known for these properties in our sample. CNTF was shown to promote survival, migration, and differentiation of OPCs into mature A p-value of .01 was considered as significant and is marked with a*. Strong correlations are marked with an underline. myelin forming cells (Modi, Sendtner, & Pahan, 2013). Furthermore, CNTF was additionally shown to promote the survival of neurons and to protect mature neurons following CNS injury (Dallner, Woods, Deller, Kirsch, & Hofmann, 2002;Linker et al., 2002;Yokota et al., 2005).
In our sample, CNTF positive astrocytes reached a peak in IA/ER of 70%; in A and ER/LR the percentage was still over 30%. The parallel peak of CNTF expression and accumulation of OPC suggests a potent communication between this factor and this cell type. Presumably the high expression of CNTF allowed better survival of OPCs in our experimental animals than in MS brains, where remyelination often fails. Induction of CNTF signaling has already been discussed as a potential therapeutic target in MS patients (Dutta et al., 2007).
Another astrocyte-derived factor mainly thought to be involved in promoting migration of OPCs is BDNF. It is a member of the nerve growth factor (NGF) neurotrophin family and known as mediator of adult neurogenesis (Watts, Mcconkey, Anderson, & Caldwell, 2005).
Furthermore, it was shown to prevent axonal and neuronal damage after various pathological insults to the brain (Gravel, Götz, Lorrain, & Sendtner, 1997;Weibel, Kreutzberg, & Schwab, 1995). In MS brain specimen, it is primarily found in immune cells and reactive astrocytes with the highest number of BDNF positive cells being found at the actively demyelinating edge early in the development of a MS lesion, most likely to protect nerve cell processes at risk of bystander damage (Stadelmann et al., 2002). In line with this assumption, in our The switch of astrocytic phenotypes A1 and A2 during lesion evolution. In all IF stainings, the A1 phenotype (C3d) appears in red and the A2 phenotype (S100A10) in green. In all IHC stainings, the A1 phenotype appears in brown and the A2 phenotype in purple. In (a) a early remyelinating lesion is shown on a IF stained slide where the A2 phenotype is mainly present and the A1 phenotype is spread over the lesion in a granular-like pattern indicated by the arrows. In (b) a IHC of a late remyelinating lesion is marked by the circle where the A2 phenotype (purple) is mainly present. The arrows point at the lesion border where the A1 phenotype (brown) accumulates. This remarkable border is also shown in another late remyelination lesion in (c) indicated by the arrows. In (d) and (e), IF stainings are shown with different lesion types side-by-side. The circles show SPs where the A2 phenotype (green) is mainly present. In the rectangles early remyelinating areas are marked with the A1 phenotype (red) spread over in a granular-like pattern. The arrows show the border of A1 phenotype, which was detectable during later remyelination steps in our sample [Color figure can be viewed at wileyonlinelibrary.com] experimental sample, we could detect an increase in BDNF positive astrocytes from A to IA/ER and a decrease during late remyelination.
The next mandatory step to a successfully remyelinated lesion is the differentiation of OPCs to mature oligodendrocytes. CXCL-12, FN1, and Tn-C are exemplary factors supporting this critical step.
Chemokines are important keyplayers in the regulation of OPCs and interact with chemokine-receptors on the OPC cell-surface. In the normal CNS CXCL-12 is constitutively expressed on endothelial cells and a small number of astrocytes, while in MS CXCL-12 levels were shown to be high on astrocytes in both active lesions and at the edges of silent lesions where repair processes were taking place (Calderon et al., 2006 In chronically demyelinated MS lesions, FN1 expression is found in the form of aggregates suggested to be resistant to degradation. Those aggregates are described in the literature to contribute to remyelination failure by inhibiting OPC differentiation (Stoffels et al., 2013). In our work, we found FN1 to be upregulated (present in around 50% of all detected astrocytes) during active demyelination as well as in later stages of remyelination. During active demyelination, it might again be necessary to prevent OPCs from too early differentiation. Whether FN1 could indeed be detrimental or not during later remyelination has to be investigated further.
Tn-C is described in the pathogenesis of CNS autoimmunity and various pro-inflammatory mediators have been shown to induce expression of Tn-C (Mom cilovi c et al., 2017). In line with this observation, we found a Tn-C peak with 50% of all astrocytes expressing this factor during active demyelination. In addition, Tn-C production by reactive astrocytes is suggested to result in glial scar formation impeding remyelination and axonal repair in MS lesions (Gutowski, Newcombe, & Cuzner, 1999). In our work, Tn-C was also present during remyelination, which may support the presence of astroglial reaction.
Similar to BMP-2, the view on this factor depends on the perception of the impact of the glial scar. However, the role of Tn-C is indeed much more complex and our investigation can only reflect a small part of it. Tn-C is, for example, further described to disturb the blood-brain barrier in an experimental mouse model and it is also upregulated and detectable in MS sera and lesions (Gutowski et al., 1999;Harada et al., 2015). In addition, it is described that Tn-C deficiency protects against EAE (Mom cilovi c et al., 2017).
Finally, OPCs must be able to survive and proliferate in the lesion milieu. IGF is one factor that is discussed as supporting this function.
In animal models of MS, an astrocytic IGF upregulation in demyelinating lesions was found (Hinks & Franklin, 1999) and in the absence of IGF signaling remyelination does not adequately occur as OPCs do not proliferate or survive (Mason, Xuan, Dragatsis, Efstratiadis, & Goldman, 2003). In our sample, numbers of IGF expressing astrocytes increased steadily from A to LR lesions where they reached their maximum. In SP, IGF expression was similar to the pattern found in NAWM. Comparing these observations to our OPC findings, we suggest a role of IGF in OPC survival, maturation, and myelin formation.
The two factors FGF-2 and IL-6 in our sample set were present in a rather uniform manner during lesion evolution and we could not detect any clear peaks. FGF-2 is a mediator of adult neurogenesis and described in the literature to be involved in the recruitment of OPCs (Watts et al., 2005). Recent research showed that FGF-2 is highly expressed by active astrocytes in the subventricular zone in the adult brain, enhances the recruitment of progenitor cells, and therefore facilitates myelin repair (Azin, Mirnajafi-Zadeh, & Javan, 2014;Huang & Dreyfus, 2016). In experimental demyelination, FGF-2 mRNA levels peaked at the initial stage of remyelination supporting a remyelinationenhancing role (Messersmith, Murtie, Le, Frost, & Armstrong, 2000).
Furthermore, in EAE FGF-2 has been identified as a neuroprotective factor preventing and even treating this disease (Woodbury & Ikezu, 2014), whereas another animal study found, on the contrary, a demyelinating effect of continuously increased levels of FGF-2 and an inhibitory effect on OPC differentiation (Zhou, Flint, Murtie, Le, & Armstrong, 2006). In line in our study, FGF-2 was expressed on astrocytes in all lesion types but rather at uniformly low levels.
IL-6 and CNTF are structurally and functionally related molecules and both belong to the family of neuropoietic cytokines (Schönrock, Gawlowski, & Brück, 2000). In MS brain specimens, IL-6 was observed in only 10-17% of the astrocytes (Schönrock et al., 2000). In contrast to this finding, in our sample, we detected a constantly expression level of 40-60% of IL-6 positive astrocytes over all lesion types.
When looking at the results of astrocytic factors, our findings suggest a very fine graduated and balanced system with different factors having to be present at particular time points to archive successful repair. It is often not possible to identify some factors as purely remyelination facilitating and some others as detrimental; rather the individual factors appear to have a distinct function at a distinct lesion stage. Recent work of Anderson et al. showed that preventing glial scar formation, attenuating scar forming astrocytes, or deleting chronic astrocyte scars in severe spinal cord injury lesion in mice failed to result in a regrowth of axons. They even found out that a loss of the glial scar prevents stimulated axon regrowth. They therefore concluded that the glial scar formation rather aids than prevents axon regeneration (Anderson et al., 2016).
In addition to the formation of a glial scar, the very complex role of astrocytes is further characterized by their ability to exhibit different phenotypes and thereby to broaden their operational area. Neuroinflammation induces two different types of reactive astrocytes that were named as "A1" and "A2", respectively (Liddelow et al., 2017). A1 astrocytes are induced by classically activated neuroinflammatory microglia and were described as harmful by inhibiting OPC proliferation and differentiation, whereas A2 astrocytes were suggested to be protective. In line with that we found A1 astrocytes mainly present in active lesions and the A2 phenotype during remyelination. Interestingly, the distribution of the A1 phenotype also changed with the start of remyelination. During early remyelination, A1 was detectable in a granular pattern spreading over the lesion. During later remyelination, the A1 phenotype formed a ring surrounding the lesion, which might acts as a barrier, protecting adjacent healthy tissue.

| CONCLUSION
From our data, we conclude that the formation of a glial scar after a