Identification of a microglia phenotype supportive of remyelination

Authors

  • Marta Olah,

    1. Department of Neuroscience, Section Medical Physiology, University Medical Center Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Sandra Amor,

    1. Department of Pathology, VU University Medical Centre, Amsterdam, The Netherlands
    2. Neuroimmunology Unit, Neuroscience Centre, Blizard Institute of Cell and Molecular Science, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    Search for more papers by this author
  • Nieske Brouwer,

    1. Department of Neuroscience, Section Medical Physiology, University Medical Center Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Jonathan Vinet,

    1. Department of Neuroscience, Section Medical Physiology, University Medical Center Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Bart Eggen,

    1. Department of Neuroscience, Section Medical Physiology, University Medical Center Groningen, Groningen, The Netherlands
    Search for more papers by this author
  • Knut Biber,

    1. Department of Neuroscience, Section Medical Physiology, University Medical Center Groningen, Groningen, The Netherlands
    2. Department of Psychiatry and Psychotherapy, Section Molecular Psychiatry, University of Freiburg, Freiburg, Germany
    Search for more papers by this author
  • Hendrikus W. G. M. Boddeke

    Corresponding author
    1. Department of Neuroscience, Section Medical Physiology, University Medical Center Groningen, Groningen, The Netherlands
    • Department of Neuroscience, Section Medical Physiology, University Medical Center Groningen, Groningen, The Netherlands
    Search for more papers by this author

Abstract

In multiple sclerosis, endogenous oligodendrocyte precursor cells (OPCs) attempt to remyelinate areas of myelin damage. During disease progression, however, these attempts fail. It has been suggested that modulating the inflammatory environment of the lesion might provide a promising therapeutic approach to promote endogenous remyelination. Microglia are known to play a central role in neuroinflammatory processes. To investigate the microglia phenotype that supports remyelination, we performed genome-wide gene expression analysis of microglia from the corpus callosum during demyelination and remyelination in the mouse cuprizone model, in which remyelination spontaneously occurs after an episode of toxin-induced primary demyelination. We provide evidence for the existence of a microglia phenotype that supports remyelination already at the onset of demyelination and persists throughout the remyelination process. Our data show that microglia are involved in the phagocytosis of myelin debris and apoptotic cells during demyelination. Furthermore, they express a cytokine and chemokine repertoire enabling them to activate and recruit endogenous OPCs to the lesion site and deliver trophic support during remyelination. This study not only provides a detailed transcriptomic analysis of the remyelination-supportive microglia phenotype but also reinforces the notion that the primary function of microglia is the maintenance of tissue homeostasis and the support of regeneration already at the earliest stages in the development of demyelinating lesions. © 2011 Wiley Periodicals, Inc.

INTRODUCTION

Traditionally, it has been assumed that in multiple sclerosis (MS), remyelination is incomplete or fails because of the decreased availability of oligodendrocyte precursor cells (OPCs) and their inability to colonize the lesions. However, recent developments in imaging and immunohistochemical techniques have made it clear that attempts to remyelinate the demyelinated lesions are much more common than initially thought (Blakemore and Franklin,2008; Blakemore and Keirstead,1999; Patani et al.,2007; Patrikios et al.,2006). Clearly, the presence of OPCs and their migratory capacity are not limiting factors for remyelination to occur. It has been proposed that while OPCs are readily recruited to lesion sites, their differentiation toward oligodendrocytes is halted by the lack of supportive cellular environment (Franklin,2002).

Neuroinflammation, a primary hallmark of MS (Nataf,2009), has long been considered to be essentially associated with tissue injury (Lassmann,2008). However, in contrast to the initial notion that causality only existed between inflammation and demyelination, research in the past few years has also suggested a role for inflammation in myelination and remyelination (Foote and Blakemore,2005; Rodriguez,2007; Setzu et al.,2006). Given the apparent association between inflammation and regenerative processes, it has been suggested that modulation of inflammatory events might prove to be the most effective means to enhance remyelination (Franklin,2002; Franklin and Ffrench-Constant,2008).

Microglia, the resident immune cells of the central nervous system (CNS), are involved in many types of inflammatory processes in the brain (Hanisch and Kettenmann,2007; Rivest,2009). However, despite the expanding knowledge regarding the biology of microglia, their role in leukoencephalopathies is still somewhat obscure (Napoli and Neumann,2009,2010; Rodriguez,2007).

Microglia are suggested to be involved in the first steps in the pathophysiological process of MS (Van der Valk and De Groot,2000); however, their actual role in demyelination and remyelination is yet unclear. Much of the uncertainty arises from depletion studies, from which it seems that demyelination is not complete in the absence of microglia/macrophages (Brosnan et al.,1981; Huitinga et al.,1990). In contrast, remyelination is also seriously delayed and affected when microglia/macrophages are not present (Kotter et al.,2005; Li et al.,2005). Genetic ablation of the two classical proinflammatory cytokines, tumor necrosis factor alpha (TNFα) and interleukin-1 beta (IL-1β), which after application in the CNS cause neuroinflammation and subsequent demyelination (Ferrari et al.,2004; Redford et al.,1995), had no effect on demyelination in the cuprizone model, but significantly impeded remyelination due to inadequate recruitment and differentiation of OPCs (Arnett et al.,2001; Mason et al.,2001). This clearly outlines the complex nature of the relationship between inflammation and remyelination. Moreover, in addition to microglia, other players of the immune response in the CNS (such as T cells) have proven to be indispensable for efficient myelination to occur (Bieber et al.,2003).

In this study, we have identified the microglia phenotype associated with CNS demyelination and remyelination following cuprizone treatment. In this model, primary demyelination and the subsequent remyelination occur with minimal contribution from peripheral immune cells (Matsushima and Morrel,2002; McMahon et al.,2002; Remington et al.,2007), allowing us to characterize the role of microglia independently of the peripheral immunity.

Surprisingly, the microglia phenotype that supported multiple aspects of remyelination (such as activation, migration, proliferation, and differentiation of OPCs) was identical to the phenotype that was already present during the demyelination phase. Thus, our data suggest that the primary function of CNS resident microglia is repair and maintenance of tissue homeostasis.

MATERIALS AND METHODS

All experiments were approved by the Institutional Animal Care and Use Committee of the University of Groningen and were in line with local and international guidelines laid down for animal experiments aiming at minimizing the number of animals used and the discomfort.

Cuprizone Diet

Cuprizone was used to induce primary demyelination in the mouse corpus callosum. The diet containing 0.2w/w % cuprizone was freshly prepared daily by thoroughly mixing bis(cyclohexanone)oxaldihydrazone (Sigma; Cat. No. C9012) and standard powder chow (abdiets; Cat. No. 2103). Animals were provided with fresh food (∼5 g of food/mouse/day) on daily basis.

Experimental Setup and Grouping of the Animals

A total of 125 C57BL/6 male mice (Harlan) were used in this study. At the start of the experiment, the mice were 8 weeks old. Based on the time course of administration of cuprizone diet and of the recovery period, the experimental groups were designed as follows (Supp. Info. Fig. 1): control group (standard powder chow, C), 2 weeks demyelination group (cuprizone diet for 2 weeks, 2WD), 5 weeks demyelination group (cuprizone diet for 5 weeks, 5WD), 1-week remyelination group (cuprizone diet for 5 weeks followed by 1 week of chow without cuprizone, 1WR), 2 weeks remyelination group (cuprizone diet for 5 weeks followed by 2 weeks of chow without cuprizone, 2WR). Each group consisted of 25 animals housed five per cage.

Isolation and FACS of Corpus Callosum Microglia

Animals were sacrificed as described in the Supporting Information. After perfusion, the brains were cut into 2- to 3-mm-thick coronal sections from which the corpora callosa were dissected under dissection microscope in ice-cold dissection medium. The dissected tissue was homogenized in the dissection medium in a glass potter. The cell suspension was incubated with anti-CD11b PE (eBioscience; Cat. No. 12-0112) and anti-CD45 FITC (eBioscience; Cat. No. 11-0451) flow cytometry antibodies for 15 min on ice and washed in PBS. Subsequently, microglia were sorted by means of a Dako-Cytomation MoFlo High-Performance Cell Sorter using the gate for living cells (forward scatter/side scatter) together with a gate set up for the characteristic CD11b high/CD45 intermediate microglia profile. Between 40,000 and 100,000 microglia per sample were directly sorted into lysis buffer.

Genome-Wide Gene Expression Analysis

Detailed description of the RNA isolation and the subsequent microarray analysis can be found in the Supporting Information Materials and Methods section. In brief, total RNA was isolated from microglia using the RNeasy Micro kit (Qiagen) according to the manufacturer's protocol. RNA quality was checked at multiple steps along the protocol (with NanoDrop and Experion RNA HighSens quality control kit; BioRad). Samples were hybridized to MouseRef-8 v2.0 Expression BeadChips (Illumina) and scanned. After quantile normalization in GenomeStudio, the raw data were analyzed with the help of GeneSpring GX software v10.0 (Agilent). For detailed description of the statistical parameters applied and the additional softwares used for data interpretation and visualization, please see the Supporting Information Materials and Methods section.

Histochemistry, Immunohistochemistry, and In Situ Hybridization

The brains of 25 animals (five animals per group) were processed to assess the extent of demyelination and remyelination following cuprizone treatment, to confirm the differential expression of genes at the mRNA and protein level in situ, and to investigate the cellular events associated with demyelination and remyelination. Changes in the level of myelination of the corpus callosum were detected with Luxol Fast Blue (LFB) histochemistry. To detect the ingestion of myelin debris by microglia, oil red O (ORO) staining was performed. The description of both histochemical methods can be found in the Supporting Information Materials and Methods section.

Immunohistochemistry on mouse tissues was performed according to standard protocols described in the Supporting Information Materials and Methods section. The antibodies are listed in the same section. In situ hybridization was applied to detect the changes in the expression level of certain selected transcripts. A detailed description of the generation of custom-made riboprobes and the in situ hybridization procedure is given in the Supporting Information Materials and Methods section.

Image Acquisition

Fluorescent images were made with a Leica SP2 AOBS confocal laser scanning microscope and Zeiss Axioskop 2 fluorescent microscope equipped with a Leica DFC300FX camera. Bright field images were acquired with Olympus BX50 research microscope equipped with a CCD camera using the SIS software.

RESULTS

Cuprizone Diet Caused Demyelination in the Corpus Callosum Followed by Rapid Remyelination on Drug Withdrawal

To investigate the microglia phenotypes associated with demyelination and remyelination, we used the cuprizone model (Matsushima and Morrell,2002). Our experimental design is shown in Supp. Info. Fig. 1A. Five weeks of 0.2% w/w % cuprizone diet (5WD) resulted in pronounced demyelinated lesions in the corpus callosum. Demyelination was determined using LFB and immunostaining for myelin basic protein (Fig. 1A,B). After withdrawal of cuprizone, fast recovery was visible already at 1WR. By the end of the second-week postcuprizone withdrawal (2WR), the staining intensity and the structural organization of the affected corpus callosum were back to the control (C) levels (Fig. 1A,B).

Figure 1.

Cellular events during demyelination and remyelination. A: Luxol Fast Blue (LFB) histochemistry was used to detect demyelinated lesions in the corpus callosum. Intact myelin in the control animals showed dense, strong blue staining in the white matter. The staining pattern did not change much in the 2WD group, whereas the 5WD group showed clear demyelinated lesions throughout the rostrocaudal extent of the corpus callosum. These lesions were characterized by the lack of myelin staining and an increase in cellularity. By the end of the first week of recovery, remyelination was almost complete. In the 1WR and the 2WR group, demyelination was detectable only in patches in the corpus callosum. B: Myelin basic protein (MBP) immunohistochemistry. Changes in the intensity and distribution of MBP immunoreactivity in the corpus callosum showed slightly different time course than of LFB histochemistry. At 2WD, MBP staining intensity was already diminished. At 5WD, MBP immunopositivity was punctate at the lesion site. Recovery was seen already at 1WR; however, even at 2WR, the staining intensity and organization did not reach the control levels. C: Iba1 immunohistochemistry. Microglia of the corpus callosum displayed a resting morphology in control condition, whereas their number increased and their morphology changed toward an activated state with occasional amoeboid cells visible in the 5WD and 2WR group (cc, corpus callosum; scale bar stands for 500 μm in [A], 100 μm in [B], and 20 μm in [C]).

Figure 2.

General overview of the gene expression analysis. A: Venn diagram depicting shared and unique genes expressed by microglia at C, 5WD, and 2WR conditions. The absence or presence of the gene transcripts in the different conditions was determined based on an arbitrary cut off value (see “Materials and Methods” and “Results” sections). Major gene ontology annotations of unique and shared genes between the C, 5WD, and 2WR groups are listed (green: present only in C; blue: present only during 5WD; red: present only during 2WR). B: Gene expression patterns across the experimental conditions. Three major groups of genes could be identified based on the pattern of the changes occurring in their expression levels during demyelination and remyelination. One group of genes was showing significant downregulation already at 2WD (when compared to C), and their expression was further downregulated at 5WD and remained downregulated during remyelination (1WR, 2WR; a′). A second group of genes showed upregulation only in 5WD, and during remyelination, their expression returned to the levels observed in C condition (b′). The third group consisted of genes that were not expressed (or marginally expressed) at the control condition, but were gradually upregulated during demyelination and remained upregulated (or were further upregulated) during remyelination (c′). C: Gene ontology (GO) analysis of the genes that showed different expression patterns across the samples. Bars represent the percentage of genes from a given gene list belonging to a given gene ontology term. Diamonds represent the −log 10 transformation of the P value associated with the level of enrichment of a given GO term with genes from our list genes. The red line represents the significance cutoff value (−log 10 of 0.05 = 1.3). (C-a′), (C-b′), and (C-c′) corresponds to (B-a′), (B-b′), and (B-c′), respectively. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 3.

Significant differently regulated microglial genes in the course of demyelination and remyelination in the cuprizone model. A: Volcano plots showing the significantly differently expressed genes between pairs of conditions (unpaired T test, P < 0.05, FC > 2). The abscissa represents the log 2-transformed fold change in gene expression levels between pairs of conditions. Negative values indicate downregulation, whereas positive values represent upregulation. The ordinate represents the −log 10 transformation of the corrected P value. Red lines represent the cutoff values for both the corrected P value and the fold change. Red diamonds correspond to genes that passed the significance testing, whereas gray diamonds symbolize the genes that did not. B: Scatter plots displaying the log 2-transformed normalized expression values (NEVs) of the significantly different expressed genes in pairs of conditions. Red lines represent the cutoff value of the fold change difference (FC > 2) and an arbitrary criterion (NEV > 3 at least at one of the conditions) aiming at excluding the marginally expressed genes. C: Venn diagrams demonstrating the number of shared genes between the pairwise comparisons and the number of genes that were unique to them. D: Bar graph showing the direction and the extent of changes occurring in the expression levels of selected genes between pairs of conditions. On the Y-axis, log 2 transformation of the fold change is depicted. Negative values indicate downregulation, whereas positive values represent upregulation. For further details on the function of the mentioned genes, see Supp. Info. Table 1 (C, control group; 5WD, 5 weeks demyelination group; 2WR, 2 weeks remyelination group). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 4.

Transcriptomic “fingerprint” of microglia in the course of demyelination and remyelination. The log 2-transformed normalized expression values of the array data (an average of three independent observations for each gene) was used to produce the heat maps. Please see the reference bar at the bottom for color coding of the expression levels; black corresponds to absent/not expressed. A:The microglia phenotype associated with demyelination and remyelination in the cuprizone model. Heat maps showing the expression levels of marker genes attributed to classically and alternatively activated macrophages, dendritic cells, and microglia. B: The effector functions of microglia during demyelination and remyelination in the cuprizone model. Heat maps showing the expression levels of genes associated with phagocytosis of apoptotic cells and myelin, tissue remodeling, cholesterol homeostasis of macrophages, secreted cytokines, and chemokines relevant for the recruitment and differentiation of OPCs. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Microglia Activation in the Corpus Callosum During Demyelination and Remyelination

Microglia in the corpus callosum of 2WD animals showed no morphological sign of activation; however, at 5WD, microglia at the lesion sites were activated (Fig. 1C). Immunohistochemistry against the microglia/macrophage marker ionized calcium-binding adaptor molecule 1 (Iba1) confirmed that most of the cells present at the lesion sites were microglia/macrophages (Figs. 1C, 5, and 7). Outside the lesions in the corpus callosum at 5WD, microglia formed a mixed population, showing morphological phenotypes ranging from ramified to amoeboid. The number of microglia and their state of activation in the corpus callosum diminished after withdrawal of cuprizone, which was evident at 1WR. By 2WR, the corpus callosum microglia regained their ramified morphology, though activated microglia were still present in patches, at sites where lesions used to be present (Fig. 1C).

Figure 5.

The presence of high levels of MHCII on microglia and the absence of costimulatory molecules suggests a tolerance-inducing phenotype. Representative photomicrographs of immunohistochemical stainings showing the gradual upregulation of MHCII on the microglia and the presence of CD3 positive T cells in the corpus callosum parenchyma during demyelination and remyelination. All images depict the corpus callosum (C, control; 5WD, 5 weeks of demyelination; 2WR, 2 weeks of remyelination; Iba1, ionized calcium-binding adaptor molecule 1, a microglia marker; MHCII, major histocompatibility complex II; CD3, a T cell marker; Hoechst, a nuclear dye; calibration bar represents 100 μm).

Figure 6.

Myelin phagocytosis by lesion-associated microglia shapes their phenotype supportive of remyelination. A: Myelin degradation products (in the form of lipid droplets) were detected in microglia with the means of ORO histochemistry. Please note the faint pink staining of the intact myelin in the control condition and at 2WD, the accumulation of microglia that ingested myelin at the lesion site at 5WD, the persistent presence of lipid droplets in microglia at 1WR, and their disappearance at 2WR. B: Representative photomicrographs showing the significant upregulation of Apoe mRNA in microglia associated with demyelination and remyelination as detected with in situ hybridization. C: Representative photomicrographs showing the significant upregulation of Lpl mRNA in microglia associated with demyelination and remyelination detected by the means of in situ hybridization (2WD, 2 weeks of demyelination; 5WD, 5 weeks of demyelination; 1WR, 1 week of remyelination; 2WR, 2 weeks of remyelination; ORO, oil red O histochemistry; Apoe, apolipoprotein E; Lpl, lipoprotein lipase; L, lesion; cc, corpus callosum; calibration bars in the main images represent 250 μm, whereas in the insets, they represent 50 μm).

Figure 7.

Confirmation of the expression patterns of selected genes at the protein level by the means of immunohistochemistry. For all the examined markers, the changes detected at the protein level were in accordance with the changes in the gene expression levels. Tgfβ (A) and Spp1 (B) were both upregulated in the course of demyelination and remained upregulated in microglia during the remyelination phase. White arrows point toward the double-positive cells in the corpus callosum. C: Antileukoproteinase (a potent anti-inflammatory secreted serine protease inhibitor) was upregulated during 5WD, but was absent from microglia at other time points. D: Some microglia at 1WR and 2WR expressed detectable levels of the dendritic cell marker, CD11c. E: Galectin 3 mRNA was gradually upregulated in microglia in the course of demyelination and remyelination, became detectable at the protein level at 5WD (data not shown), and remained upregulated up to 2WR (5WD, 5 weeks of demyelination; 1WR, 1 week of remyelination; 2WR, 2 weeks of remyelination; Tgfβ, transforming growth factorbeta; Spp1, osteopontin; Iba1, ionized calcium-binding adaptor molecule 1; IB4, isolectin B4; Slpi, antileukoproteinase; Cd11c, integrin alpha-X, a dendritic cell marker; Gal3, galectin-3; calibration bars in [A] and [B] main images stand for 100 μm, whereas in the insets of [A] and [B], they represent 50 μm; in [C]–[E], the scale bar stands for 50 μm).

Microarray Gene Expression Analysis Revealed Distinct Gene Expression Patterns in Microglia in the Course of Demyelination and Remyelination

We aimed to determine gene expression in microglia, specific to the different experimental groups (C, 5WD, and 2WR). Microglia in the corpus callosum showed the expression of 7,500, 9,000, and 9,000 genes under physiological conditions, during demyelination, and in the course of remyelination, respectively, of which around 6,200 genes were shared among all three conditions (Fig. 2A). Apart from genes present at all three conditions, few genes were shared by C and 5WD (55 genes; 0.7% in C and 0.6% in 5WD) and by C and 2WR (147 genes; 1.9% in C and 1.6% in 2WR). In contrast, the genes shared between 5WD and 2WR (2,185 genes) represented ∼22% of the total expressed genes.

Detailed analysis of the microarray data revealed three distinct patterns in gene expression of demyelination- and remyelination-associated microglia (Fig. 2B). One group of genes showed rapid downregulation during demyelination and remained downregulated during remyelination (Fig. 2B-a′). A second group of genes was upregulated in the course of demyelination and remained upregulated after remyelination was complete (Fig. 2B-c′). A third group of genes was induced in the 5WD group, but returned to control levels on cuprizone withdrawal (Fig. 2B-b′).

Subsequently, we performed gene ontology (GO) analysis. Around 20% of genes of the first group (Fig. 2B-a′) was annotated to the ontology term “cholesterol metabolic process” (GO:0008203; Fig. 2C-a′). Within the first group, 15% of the genes belonged to the ontology term “acute inflammatory response” (GO:0002526; Fig. 2C-a′). Of the second group (Fig. 2B-b′), almost 25% of genes were associated with the GO term “cell cycle” (GO:0007049; Fig. 2C-b′). From the genes belonging to the third group (Fig. 2B-c′), around 20% belonged to the GO term “immune response” (GO:0006955; Fig. 2C-c′). Another significant GO term in the third group was “antigen processing and presentation” (GO:0002478; Fig. 2C-c′), which contained more than 20% of the genes. The lists of genes belonging to the three groups described above are summarized in Supp. Info. Table 1.

Statistical Analysis of the Microarray Data Revealed Differently Expressed Microglia Genes During Demyelination and Remyelination

To explore genes that were differentially regulated, we conducted comparisons between pairs of conditions (C vs. 5WD, C vs. 2WR and 5WD vs. 2WR). Comparison of C versus 5WD and C versus 2WR yielded both around 500 genes that were significantly upregulated and more than 200 genes that were significantly downregulated (Fig. 3A). In the comparison of 5WD to 2WR, ∼30 genes were significantly upregulated and around 50 genes were significantly downregulated (Fig. 3A). Figure 3B shows the actual expression levels of the significantly different expressed genes in pairwise comparisons. The examples clearly demonstrate that the direction and the extent of changes in gene expression levels are similar between C versus 5WD and C versus 2WR, which differed from the pattern in gene expression changes found in 5WD versus 2WR. Overall, ∼60% of genes that were significantly upregulated from C to 5WD and from C to 2WR were shared by these two comparisons (Fig. 3C). The percentage of shared downregulated genes was also similar (Fig. 3C). In contrast, the significantly regulated genes between the conditions 5WD and 2WR were not shared with the other two comparisons.

The examples given in the volcano plots (Fig. 3A) demonstrate and confirm the changes of the gene expression observed at the initial rough analysis (Fig. 2). Most genes that were upregulated/downregulated from C to 5WD remained significantly upregulated/downregulated during remyelination (e.g., H2-Aa and cholesterol 25-hydroxylase [Ch25h] for upregulated genes and Hmgcs2 for downregulated genes; see Fig. 3A,B). In Fig. 3D, we depicted the extent and directionality of changes in expression levels of certain example genes between the three comparisons. In Fig. 3D-a′ and -b′, genes are arranged according to the descending order of their fold change in the 5WD versus 2WR and C versus 5WD comparison, respectively. It is evident (Fig. 3D-b′) that the pattern of gene expression changes between C and 5WD and that C and 2WR is largely similar. The few significant changes that occur during the transition from demyelination to remyelination are either due to the further upregulation of some immunity-associated genes (e.g., H2-Aa; Fig. 3D-a′) or downregulation of cell proliferation-associated genes (e.g., Birc5 and Kif22; Fig. 3D-a′) that were transiently induced during the demyelination phase.

Expression of Markers of Microglia/Macrophage Activation Reveals the Transcriptomic “Fingerprint” of a Remyelination-Supportive Microglia Phenotype

The analysis of the microglia gene expression in our model (Figs. 2 and 3) points toward the presence of a single remyelination-supportive microglia phenotype that develops during the course of demyelination.

Microglia are believed to adapt a suppressed immunophenotype under physiological conditions and to obtain a macrophage-like phenotype on activation. Therefore, we assembled a list of microglia/macrophage activation markers as well as genes associated with microglia/macrophage effector functions based on available literature. Expression patterns of genes characteristic of certain microglia/macrophage phenotypes (Fig. 4A) and microglia genes associated with tissue homeostasis and regeneration (Fig. 4B) are depicted in heat maps. On the basis of the microarray data, we found that microglia during demyelination and remyelination upregulated some markers of M1-type macrophages (e.g., Tlr4, Lyz1 and Lyz2, Tnf, Il1a and Il1b, and Mmp12; Fig. 4A). A similar pattern was found for markers of M2-type macrophages (e.g., Il13ra1, Il4ra, and Il2rg were increasingly expressed during demyelination and remyelination). Most prominent was the upregulation of genes associated with major histocompatibility complex II (MHCII; e.g., Cd74, H2-Aa, and H2-Ab1) during demyelination and remyelination. The co-stimulatory molecules (list generated based on Viglietta and Khoury,2007) were either not expressed (e.g., Cd40 and Cd80) or not regulated (e.g., Cd86). Interestingly, the only co-stimulatory molecule that was upregulated in a similar fashion as the MHC genes was Cd274. Some dendritic cell (DC) markers were also induced during demyelination and remained upregulated during remyelination (e.g., CD83, Clec7a, and Itgax; Fig. 4A). Most microglia markers (Aif1 [Iba1], Emr1 [cell surface glycoprotein F4/80], and Sfpi1 [transcription factor PU.1]) were expressed ata stable level across the samples, with only a mild downregulation of Cx3cr1 and P2ry12 on activation. Based on the identity of differentially regulated genes, the following microglial effector functions in demyelination and remyelination are proposed: phagocytosis of apoptotic cells and myelin debris (e.g., Lrp1, Calr, Cd14 and Itgb2, Itgam, and Lgals3), salvage of myelin constituents (e.g., Hmgcs2, Lpl, and Apoe), recruitment of OPCs and trophic support for the remyelinating oligodendrocytes (e.g., Cxcl10, Cxcl13 and Igf1, Tgfb1, Pdgfa, and Pdgfb), and tissue remodeling (e.g., Mmp12 and Mmp14).

Confirmation of Gene Expression Analysis Findings In Situ at mRNA and Protein Level

We confirmed the findings of our gene expression study by investigating the expression levels of selected genes in situ at mRNA or protein level. We have primarily focused on markers associated with the state of microglia activation (tolerance inducing and anti-inflammatory) and markers indicative of involvement of microglia in regenerative processes (trophic and metabolic support to (re)myelinating oligodendrocytes). Allinvestigated markers showed an expression pattern at the protein level identical to its mRNA expression pattern.

Under control conditions, no expression of MHCII in the CNS other than in macrophages associated with the ependymal ventricular lining was observed. At 5WD, few microglia within and outside the lesions expressed MHCII in the corpus callosum. The number of MHCII positive microglia robustly increased during remyelination (Fig. 5). Because MHCII is involved in crosstalk between microglia/macrophages and T cells (Carson,2002), we investigated whether T cells were present in the corpus callosum during demyelination and remyelination. We performed immunohistochemistry against CD3, a pan T cell marker. In control animals, no T cells were found in brain parenchyma (Fig. 5). At 5WD and 2WR, a few CD3 positive T cells were present in the corpus callosum. Next, we tested the expression of the co-stimulatory molecules CD40, CD80, and CD86 on microglia at protein level. In line with our array study, the expression level of none of these markers was above the detection limit of immunohistochemistry at any investigated time point (data not shown).

Data from our gene expression study suggest that microglia might recycle cholesterol and lipids from degenerating myelin and make it available for remyelinating oligodendrocytes. To detect myelin degradation products in microglia, we performed ORO histochemistry in brain sections, which showed that microglia in the corpus callosum readily scavenge the myelin debris. At 5WD, microglia at the lesion site contained numerous lipid droplets (Fig. 6A), which disappeared after 2 weeks.

Subsequently, we analyzed the expression of two factors involved in lipid recycling, namely apolipoprotein E (Apoe) and lipoprotein lipase (Lpl), which were significantly regulated in our array study. Their expression levels were investigated by in situ hybridization. ApoE mRNA was robustly upregulated in microglia in the corpus callosum at 5WD and remained upregulated at 2WR when compared with C (Fig. 6B). In control brains, we did not detect any mRNA signal for Lpl (Fig. 6C). However, lesion-associated microglia at 5WD showed strong positivity for Lpl and retained the expression level of this enzyme at 2WR.

We also investigated protein expression of transforming growth factor beta-1 (TGF-β1), osteopontin, antileukoproteinase (ALP), CD11c, and galectin-3, which all showed significant regulation during demyelination and remyelination. TGF-β1 expression was upregulated in the course of demyelination and remyelination in microglia both at mRNA (Tgfb1; Fig. 4B) and protein levels (Fig. 7A). Osteopontin (Spp1) was detected in microglia at 5WD as well as in microglia in the remyelinating corpus callosum (Fig. 7B); however, it was absent in white matter microglia of control animals. Secretory leukocyte peptidase inhibitor (Slpi) or ALP has been implicated in the resolution of inflammation and wound healing (Ashcroft et al.,2000; Doumas et al.,2005; Odaka et al.,2003). ALP immunoreactivity in microglia was only detectable at 5WD in the corpus callosum (Fig. 7C). Some microglia in the corpus callosum expressed the DC marker CD11c (Itgax; Fig. 7C), and this became particularly apparent during remyelination (1WR and 2WR). Expression of galectin-3 (Lgals3; Fig. 4B), a galactose-binding lectin mainly associated with microglial myelin phagocytosis (Reichert and Rotshenker,1999; Rotshenker et al.,2008), was already detectable at 5WD on microglia in the corpus callosum and remained upregulated during remyelination (1WR and 2WR; Fig. 7E).

DISCUSSION

In MS lesions, remyelination fails due to the lack of regeneration-supportive cellular environment (Franklin,2002). Given the intimate and complex relationship between inflammation and remyelination (Hohlfeld et al.,2007; Rodriguez,2007; Sharief,1998), it has been suggested that modulation of the local inflammatory milieu might prove to be the most promising therapeutic approach to support remyelination in MS lesions (Franklin and Ffrench-Constant,2008). Nonetheless, any enterprise to devise therapeutic interventions that would aim at establishing or ameliorating the remyelination-supportive cellular environment in the CNS of patients with MS has been greatly hindered by the lack of knowledge regarding the exact nature of this cellular environment. To fill this void, we set out to investigate the demyelination- and remyelination-supportive phenotypes of microglia—the resident immune cells of the CNS that govern all neuroinflammatory processes.

In our study, we determined the global gene expression changes of microglia during demyelination and remyelination by means of microarray analysis, followed by detailed confirmation of our findings at the protein level. We used the cuprizone model in which the uncompromised blood–brain barrier (Bakker and Ludwin,1987; Kondo et al.,1987; McMahon et al.,2002) and the lack of infiltrating peripheral immune cells (McMahon et al.,2002; Remington et al.,2007) allow detailed study of the microglia phenotype associated with demyelination and remyelination. Moreover, demyelination and remyelination in the cuprizone model follow a highly reproducible time course (Gudi et al.,2009; Matsushima and Morell,2002), making it technically feasible to experimentally dissect the two processes and the related microglia phenotype changes.

Surprisingly, we did not find evidence supporting the idea that there would be two microglia phenotypes—one associated with demyelination and another specifically responsible for remyelination. Clearly, the microglia phenotype present during remyelination developed gradually during the course of demyelination and persisted on remyelination. Thus, most of the genes that were regulated showed a unidirectional upregulation or downregulation throughout the demyelination and remyelination process.

One group of genes that most prominently showed this robust expression pattern in the course of demyelination and remyelination was associated with the MHCII complex (e.g., H2-Aa, H2-Ab1, H2-Eb1, and Cd74). Although many of the MHC molecules were strongly induced with expression levels peaking at 2WR, the lack of regulation of the co-stimulatory molecules was remarkable. Cd86 (B7.2) was present in microglia, but its expression levels were not regulated; however, Cd80 (B7.1) was not expressed at any time point. A similar phenotype has been described in an acute experimental autoimmune encephalitis (EAE) model by Almolda et al. (2010), in which MHCII and CD86, but not CD80, were strongly induced in microglia and were suggested to be involved in lymphocyte inhibition and tolerance induction. Importantly, similarly to our study, they also found this phenotype to persist in the recovery phase. Interestingly, the only co-stimulatory molecule being strongly regulated during demyelination and remyelination was Cd247 (B7-H1), a known inhibitory co-stimulatory molecule (Keir et al.,2008; Magnus et al.,2005; Ortler et al.,2008). Its expression was already induced at 2WD and remained upregulated during remyelination. Importantly, Duncan and Miller (2011) also found early upregulation of B7-H1 on microglia in Theiler's murine encephalomyelitis virus (TMEV)-induced demyelinating disease, another mouse model of MS. Moreover, in their study, in vivo blockage of B7-H1 significantly exacerbated the disease, suggesting an active role of microglial B7-H1 expression in limiting T-cell-mediated immune responses in the CNS.

The notion that the microglia phenotype that governs demyelination and remyelination in this model is a tolerogenic one is further supported by the intriguing expression pattern of certain DC markers on microglia. Next to Itgax (CD11c) that was moderately induced during demyelination and remyelination, we observed a strong induction of Cd83, which is known to be associated with the tolerogenic DC phenotype (Fujimoto and Tedder,2006). We also found Clec7a (also known as dectin-1) to be strongly induced in microglia. Interestingly, signaling through this particular C-type lectin has been shown to be a negative regulator of microglia activation (Shah et al.,2009).

Importantly, we did not find upregulation of classical inflammatory mediators other than IL-1β and TNFα in our study, both of which have been shown to be indispensable for remyelination to occur (Arnett et al.,2001; Mason et al.,2001). Interestingly, inducible nitric oxide synthase (iNOS) was found to be induced in the cuprizone model in microglia (Arnett et al.,2002; Liñares et al.,2006) and was moderately protective against demyelination (Arnett et al.,2002). Nonetheless, iNOS was not among the genes expressed in our study. The discrepancy with previous studies might stem from the differences between the time points when iNOS expression was examined (3.5 weeks of demyelination vs. 5 weeks of demyelination in our case).

In line with previous studies, we found strong induction of cystatin-F (Cst7) as early as 2WD, which persisted throughout remyelination. Since Ma et al. (2011) found a positive correlation between microglia cystatin-F expression and the regenerative capacity of the CNS tissue in multiple MS models and MS spinal cords as well, the expression profile of this particular enzyme in our study further supports our claim that the principal function of microglia is support of regeneration.

One aspect of the described microglia phenotype is its trophic support of oligodendrocyte differentiation. We found upregulation of Pdgfa, Pdgfb, Vegfa, Vegfb, Tgfb1, Igf1, and Spp1, all of which have been reported to promote oligodendrocyte differentiation (Baumann and Pham-Dinh,2001; Bradl and Lassmann,2010; Chesik et al.,2008; Diemel et al.,2003; Frost et al.,2003; Hinks and Franklin,1999; Hsieh et al.,2004; Kim et al.,2009; Selvaraju et al.,2004; Vela et al.,2002; Woodruff et al.,2004). Interestingly, we did not find upregulation of Gdnf or Fgf2, as reported by Gudi et al. (2011). In their study, Gdnf induction in microglia peaked at 1 week of cuprizone treatment after which it abruptly declined, whereas the highest expression of Fgf2 was observed at 3.5 weeks of demyelination. Given the differences in the experimental setups, we might have missed these expression peaks due to the different time points examined.

Interestingly, we found evidence of another aspect of microglial trophic support of remyelination. Macrophages associated with peripheral nerve regeneration scavenge and reutilize sequestered cholesterol, offering it to remyelinating Schwann cells (Jurevics et al.,1998). We found indications that microglia play a similar role in remyelination of the CNS. Several known lipoprotein receptor mRNAs (including Sorl, Lrp1, and Scarf2) were upregulated throughout the demyelination and remyelination processes. Interestingly, a marked downregulation of genes associated with de novo cholesterol synthesis (e.g., Hmgcs2) was observed, whereas other genes associated with hydrolysis of lipoproteins (e.g., Aadacl, Lpl, and its cofactor Apoc2) were distinctly upregulated. The upregulation of Ch25h also suggests that microglia are actively repressing their endogenous cholesterol biosynthetic enzymes. Furthermore, genes associated with the cholesterol efflux system of macrophages (e.g., Abcg1, Abca7, and Apoe) were upregulated throughout demyelination and remyelination.

The gene expression data further suggest that corpus callosum microglia possess the enzyme repertoire required for tissue remodeling under demyelinating and remyelinating conditions. Among the upregulated genes were the membrane-type matrix metalloproteinase 14 and macrophage metalloellastase (Mmp14 and Mmp12, respectively). Mmp12 expression gradually increased and peaked at 2WR, whereas Mmp14 was already strongly induced at 2WD and remained upregulated till 2WR. In line with our findings, Mmp12 has also been found to be robustly upregulated in the cuprizone model during the remyelination phase by Skuljec et al. (2011) and in the late phase of TMEV-induced demyelination (Ulrich et al.,2006).

An interesting observation is that there was very little overlap between the expression patterns of microglia genes that were significantly differentially regulated in the course of demyelination and remyelination in our study and the subventricular zone (SVZ)-resident microglia gene expression profile during acute and chronic EAE (Starossom et al.,2011). Surprisingly, in SVZ microglia, the niche-supporting genes (e.g., C3, Jag1, and Il18bp) displayed strong induction during the acute phase and their expression was diminished during the chronic phase. None of these molecules were regulated in corpus callosum microglia during demyelination and remyelination. This observation emphasizes the concept that regional differences might exist in the phenotype and responsiveness of microglia cells and that most likely this is particularly true for microglia residing in specialized niches such as SVZ.

The question that what elicits the development of the regeneration supportive microglia phenotype in the cuprizone model remains to be investigated. One possibility is that the emergence of the remyelination-supportive microglia phenotype is causally linked to one of the prototypical microglia effector functions, namely phagocytosis. Undoubtedly, clearance of tissue debris is indispensable for regeneration to occur (Kotter et al.,2006). Nonetheless, it has been shown that phagocytosis of apoptotic cells and myelin debris renders microglia/macrophages supportive of regenerative processes (Bechmann et al.,2001; Boven et al.,2006; Savill et al.,2002). Based on the wide abundance of ORO positive microglia in the corpus callosum in our model and the significant upregulation of lysozyme (a marker for phagocytic microglia; Morell et al.,1998), we propose that ingestion of tissue debris might be one of the triggers for the development of the microglia phenotype in question. It remains to be elucidated, however, whether the microglia phenotype induced by the ingestion of myelin and apoptotic cells is truly supportive of remyelination, or simply permissive. This issue could be addressed by depletion of microglia after most of the tissue debris has been removed, but before the occurrence of the large-scale remyelination. Additionally, other factors (e.g., signaling of apoptotic oligodendrocytes, recruited neural precursors, and denuded axons and/or astrocytes) may also play a role in the induction of a supportive microglia phenotype.

Several studies have applied microarray analysis to investigate gene expression changes associated with demyelination and remyelination (Arnett et al.,2003; Jurevics et al.,2002; Morell et al.,1998). These studies were, however, to a certain extent tainted by the use of whole tissue lysates. Likely, low abundance of microglial transcripts in relation to the total tissue mRNA content may have obscured the supportive microglia phenotype described here. To our knowledge, our study is the first in which changes in gene expression at genome-wide scale in microglia during demyelination and remyelination have been investigated. Clearly, the expression patterns of the markers that have been tested at the protein level matched the expression patterns on the microarray. Thus, we have strong confidence that the gene expression data reflect the microglia phenotype associated with demyelination and remyelination. Our results provide a solid ground for further studies aiming at devising means to restore the regenerative potential of the CNS tissue in MS. Moreover, it strongly supports the concept that the primary function of microglia is tolerance induction and support of regeneration.

Acknowledgements

The work of K. Biber was supported by German Research Council (DFG; FOR 1336).

Ancillary