Remodeling of the interstitial extracellular matrix in white matter multiple sclerosis lesions: Implications for remyelination (failure)

The extracellular matrix (ECM) provides protection, rigidity, and structure toward cells. It consists, among others, of a wide variety of glycoproteins and proteoglycans, which act together to produce a complex and dynamic environment, most relevant in transmembrane events. In the brain, the ECM occupies a notable proportion of its volume and maintains the homeostasis of central nervous system (CNS). In addition, remodeling of the ECM, that is transient changes in ECM proteins regulated by matrix metalloproteinases (MMPs), is an important process that modulates cell behavior upon injury, thereby facilitating recovery. Failure of ECM remodeling plays an important role in the pathogenesis of multiple sclerosis (MS), a neurodegenerative demyelinating disease of the CNS with an inflammatory response against protective myelin sheaths that surround axons. Remyelination of denuded axons improves the neuropathological conditions of MS, but this regeneration process fails over time, leading to chronic disease progression. In this review, we uncover abnormal ECM remodeling in MS lesions by discussing ECM remodeling in experimental demyelination models, that is when remyelination is successful, and compare alterations in ECM components to the ECM composition and MMP expression in the parenchyma of demyelinated MS lesions, that is when remyelination fails. Inter‐ and intralesional differences in ECM remodeling in the distinct white matter MS lesions are discussed in terms of consequences for oligodendrocyte behavior and remyelination (failure). Hence, the review will aid to understand how abnormal ECM remodeling contributes to remyelination failure in MS lesions and assists in developing therapeutic strategies to promote remyelination.

process of repair, including regeneration. For example, a role of the dynamics and the distinct involvement of ECM components is becoming increasingly apparent in the demyelinating disease multiple sclerosis (MS), showing an association between ECM alterations and white matter MS lesion formation (Satoh, Tabunoki, & Yamamura, 2009) and progression of the neuropathological state (Bonneh-Barkay & Wiley, 2009). ECM remodeling is tightly regulated by an interplay between several proteins and enzymes of which the family of matrix metalloproteinase (MMP) is a prominent proteolytic system in the spatiotemporal regulation of the ECM (Lu, Takai, Weaver, & Werb, 2011;Page-McCaw, Ewald, & Werb, 2007). Functional dysregulation of these enzymes contributes to the pathogenesis and progression of several inflammatory demyelinating diseases, including MS (Kieseier, Seifert, Giovannoni, & Hartung, 1999). Here, we aim to unravel abnormal ECM remodeling in MS lesions and link this to remyelination failure, as to assist in developing therapeutic strategies to promote remyelination in MS.
To this end, we uncover abnormal ECM remodeling in MS lesions by describing and comparing the transient-altered expression patterns of individual ECM molecules and MMPs upon successful CNS remyelination in rodent models of toxin-induced demyelination, and upon remyelination failure in MS lesions. In addition, we discuss how the (transiently) expressed ECM proteins regulate the behavior of cells that produce myelin, that is oligodendrocytes, and how their (persistent) expression may contribute to remyelination (failure). Also, the interactions between ECM molecules and MMPs and potential mechanisms leading to incorrect ECM remodeling in MS are reviewed. Before discussing this in more detail, we first present a brief overview of the pathology of MS.

| Pathological hallmarks
MS is a neurodegenerative inflammatory disease of the CNS. It is one of the most common demyelinating CNS diseases with an incidence of approximately 0.1% worldwide, while the prevalence varies based on geographical and ethnical differences (Compston & Coles, 2008;Rosati, 2001). Variable patterns of the clinical course of MS are observed. The most common form is relapsing-remitting MS, which is diagnosed in about 80% of the MS patients (Lublin & Reingold, 1996). Relapsing-remitting MS is characterized by relapses followed by full or partial recovery at the onset of the disease, but incomplete recovery arises over time and the majority of patients develop secondary progressive MS with minor remissions (Compston & Coles, 2008). Primary progressive MS is diagnosed in 10%-20% of the patients (Lublin & Reingold, 1996), has an onset at a later age, and is characterized by gradual accumulation of neurological deficits, starting already at the onset of the disease (Compston & Coles, 2008).
The disease is characterized by chronic and progressive loss of myelin sheaths surrounding the axons in the brain and spinal cord.
The primary causative mechanism(s) resulting in demyelination and progression of MS is (are) still unclear. However, genetic predisposition, that is mainly genes implied in (cell-mediated) immunity, and several environmental factors, such as vitamin D deficiency and viral infections, appear to play an important role in the development of MS (Compston & Coles, 2008;Correale & Gaitan, 2015;Olsson, Barcellos, & Alfredsson, 2017;Roch et al., 2017). These factors contribute to the inflammatory process occurring in MS, which is associated with disruption of the BBB. Whether this inflammation is a primary or secondary event in the pathogenesis of MS is unknown (Compston & Coles, 2008;Stys, Zamponi, van Minnen, & Geurts, 2012). An autoimmune response is initiated in the periphery through (as yet unknown) processes. Molecular mimicry might underlie such an event, resulting in activation of autoreactive T cells against self-antigens, which migrate to the CNS. Subsequently, this might initiate an immune response, thus causing damage and degeneration. In contrast, cytodegenerative processes in the CNS through (as yet equally unknown) factors may activate autoreactive T cells by presentation of self-antigens and subsequently induce a secondary immune response. In MS, these autoreactive T cells are directed to self-antigens that include specific proteins, present on the surface of mature oligodendrocytes and/or in myelin (Mallucci, Peruzzotti-Jametti, Bernstock, & Pluchino, 2015). This results in (additional) demyelination due to the destruction of myelin and/or mature oligodendrocytes.
The formation of white matter MS lesions is a dynamic process resulting in interlesional heterogeneity. More specifically, distinct lesions are histopathologically classified in inflammatory and demyelinating activity (Kuhlmann et al., 2017;van der Valk & De Groot, 2000). While different classification systems have been described, mainly three distinct lesions can be distinguished: active, mixed active/inactive, and inactive lesions (Figure 1, (Kuhlmann et al., 2017)).
Active lesions are the early demyelinating phenotype and are most frequently found in acute MS patients with a very short disease duration (Frischer et al., 2015), while their proportion (approx. 25%) is similar upon longer disease duration and severity (Frischer et al., 2015;Luchetti et al., 2018). Active lesions are defined by indistinct margins, harbor inflammatory activity, and contain a hypercellular lesion center with hypertrophic astrocytes, (myelin-laden) microglia/macrophages, and lymphocytes. Mixed active/inactive lesions (also called "chronic active lesions" or "smoldering lesions") show intralesional heterogeneity having a sharp border and consist of a hypocellular demyelinated lesion center with fibrous astrocytes.
The lesion center is surrounded by a broad hypercellular inflammatory rim that contains microglia/macrophages and reactive astrocytes. The mixed active/inactive lesions are more frequently observed in progressive MS patients than in relapsing-remitting MS patients (Frischer et al., 2015;Luchetti et al., 2018). The final lesion classification is a (chronic) inactive MS lesion. These lesions have a hypocellular demyelinated center, containing mainly reactive astrocytes and show hardly signs of infiltration of microglia/macrophages or lymphocytes. Inactive lesions are also predominating in progressive MS (Frischer et al., 2015).

| Failure of remyelination
Myelin is an insulating layer around axons that provides axonal protection and electrical isolation, and mediates saltatory conduction of action potentials. The loss of activity due to demyelination is partially compensated by a redistribution of sodium channels along the demyelinated parts of the axon, which allows for nonsaltatory conduction with reduced velocity (Felts, Baker, & Smith, 1997). However, this compensatory effect is only temporal and in conjunction with the loss of axonal protection, persistent demyelination leads to axonal damage and degeneration (Compston & Coles, 2008;Funfschilling et al., 2012;Lee et al., 2012). The rate of impulse transduction is reduced or the impulses cease, which results in clinical signs and symptoms of MS that reflect the affected area of the CNS (Compston & Coles, 2008). In addition, accumulation of axonal degeneration is the main process that contributes to progression of neurological dysfunction and disease severity (Compston & Coles, 2008;Papadopoulos, Pham-Dinh, & Reynolds, 2006). Progressive axonal loss is key to the continuous and irreversible neurological decline in progressive MS (Trapp & Nave, 2008). Next to primary axon damage, a major cause of axonal loss in chronic stages of MS is secondary neurodegeneration as a consequence of remyelination failure (Irvine & Blakemore, 2008). Indeed, in addition to ensure saltatory axonal conduction, myelinating oligodendrocytes secrete metabolic and trophic factors that maintain the integrity and survival of axons (Funfschilling et al., 2012;Lee et al., 2012). Therefore, prevention of axonal degeneration might be beneficial to resolve the functional deficits and progression of MS.
A regenerative process to restore myelin is required to ensure the survival of demyelinated axons. This process is called remyelination, which re-establishes saltatory conduction, protects axons from degeneration and improves clinical features of MS ; Irvine & F I G U R E 1 Schematic overview of the inter-and intralesional heterogeneity of cells, extracellular matrix (ECM) proteins, and matrix metalloproteinases (MMPs) in active, mixed active/inactive and inactive white matter multiple sclerosis lesions. Lesions are indicted as pale (demyelinated) areas. ECM proteins and MMPs depicted in green have an increased expression and ECM proteins and MMPs depicted in red have a decreased expression compared to control white matter at the indicated location of the lesion, that is lesion center or rim. COL-V, collagen V; CSPGs, chondroitin sulfate proteoglycans; Fn, fibronectin; HA, hyalorunan; OPN, osteopontin; Tn-C, tenascin-C; Tn-R, tenascin-R; Vn, vitronectin. Thrombospondin-1 expression is increased in mixed active/inactive and inactive lesions, while data on localization are not available yet [Color figure can be viewed at wileyonlinelibrary.com] Blakemore, 2008;Jeffery & Blakemore, 1997;Smith, Blakemore, & McDonald, 1979). The myelin sheaths generated in the process of remyelination are shorter and thinner, compared to myelin sheaths produced during developmental myelination, but these newlyformed sheaths suffice for axonal protection and improved functioning (Kornek et al., 2000). Experimental demyelination in rodent models show that successful remyelination is not executed by pre-existing mature oligodendrocytes but achieved by newlyformed oligodendrocytes generated from local oligodendrocyte progenitor cells (OPCs; Zawadzka et al., 2010), while OPCs present in the adult subventricular zone contribute predominantly to remyelination of lesions in their proximity (Menn et al., 2006). In response to toxin-induced demyelination in these models, astrocytes and microglia are activated (reviewed by Franklin et al., 2002). These changes lead to the (transcriptional) activation of OPCs, resulting in morphological changes and enhanced gene expression of factors involved in oligodendrocyte differentiation and maturation (Arnett et al., 2004;Ferent, Zimmer, Durbec, Ruat, & Traiffort, 2013;Moyon et al., 2015;Reynolds et al., 2002;Watanabe, Hadzic, & Nishiyama, 2004). In addition to OPC activation, microglia and astrocytes recruit and mediate the migration of OPCs to the demyelinated areas, where they further proliferate Franklin et al., 2002;Levine & Reynolds, 1999). The final essential step is the generation of new myelin sheaths and involves the differentiation of OPCs into mature oligodendrocytes. The process includes contact between the oligodendrocyte and the axon, upregulation of myelin-specific genes and generation and compaction of the myelin membranes Franklin et al., 2002). While in experimental rodent models remyelination is executed by OPCs, carbon 14-based birth dating and single nuclei RNA sequencing of MS brain tissue reveal that remyelination in MS is most likely not executed by OPCs, but by pre-existing mature oligodendrocytes (Jakel et al., 2019;Schirmer et al., 2019;Yeung et al., 2019). Whether this is an inherent capacity of remyelination in humans, or an adaptation as of the inability of OPCs to differentiate to mature myelinating oligodendrocytes in MS lesions remains to be determined. However, in both cases, that is remyelination by newly-formed or pre-existing OLGs, new myelin membranes are formed.
A plethora of molecules play an important role in the different phases of remyelination, as well as several mediators of the inflammatory response, featuring in MS Franklin et al., 2002;Gaesser & Fyffe-Maricich, 2016;Hanafy & Sloane, 2011;Miron, 2017). Although remyelination is a natural response to demyelination in most cases, this regenerative process often fails in chronic and progressive MS (Compston & Coles, 2008;Goldschmidt, Antel, Konig, Bruck, & Kuhlmann, 2009;Kuhlmann et al., 2008;Lucchinetti et al., 1999;Luchetti et al., 2018). In a subset of lesions, insufficient migration and/or proliferation of OPCs likely accounts for remyelination failure. However, remyelination mainly fails due to defective OPC differentiation Franklin et al., 2002;Kuhlmann et al., 2008). In fact, in approx. 70% of both active and chronic MS lesions, OPCs are abundantly present (Chang, Tourtellotte, Rudick, & Trapp, 2002;Goldschmidt et al., 2009;Kuhlmann et al., 2008;Lucchinetti et al., 1999;Strijbis, Kooi, van der Valk, & Geurts, 2017;Wolswijk, 1998). Of interest in this respect is that single nuclei RNA sequencing of MS brain tissue shows that more dysregulated genes are found in mature oligodendrocytes rather than in OPCs compared to tissue of healthy subjects (Schirmer et al., 2019). This may indicate that OPCs are indeed transcriptionally relatively quiescent in MS, likely as a reflection of the inhibitory environment in MS lesions for OPC differentiation. Remyelination not always fails in MS, given the presence of partly or completely remyelinated shadow plaques.
Remyelinated shadow plaques are present at all stages of MS (Frischer et al., 2015;Goldschmidt et al., 2009;Kuhlmann et al., 2017;Luchetti et al., 2018) and is thus likely executed by pre-existing oligodendrocytes (Jakel et al., 2019;Yeung et al., 2019), but the frequency of remyelinated white matter shadow plaques is lower in the progressive forms compared to relapsing-remitting MS (Luchetti et al., 2018). This is likely a reflection of remyelination failure in mixed active/inactive lesions that are dominating at later stages. Notably, although extensive remyelination is occasionally observed at late-stage progressive MS (Patani, Balaratnam, Vora, & Reynolds, 2007;Patrikios et al., 2006), it is most prominently noted in acute lesions, that is immediately upon demyelination (Prineas et al., 1989;Raine & Wu, 1993).
Remyelination failure is thus thought to be a consequence of perturbations in the different phases of remyelination, that is activation, recruitment, and differentiation of OPCs, in which aging also plays an important role Goldschmidt et al., 2009;Sim, Zhao, Penderis, & Franklin, 2002). The signaling and cellular environment, established by the state of the disease and lesion regulating these phases, is a crucial factor. Remyelination failure in MS is likely due to the presence of inhibitory signals or the lack of stimulatory signals in the damaged area. Indeed, various factors in the signaling environment of MS lesions are dysregulated, and together with ensuing cellular changes contribute to the failure of remyelination Franklin et al., 2002;Hanafy & Sloane, 2011;Miron, 2017;Williams, Piaton, & Lubetzki, 2007). In this regard, a role of the dynamics and distinct interstitial ECM components in remyelination failure in MS lesions is becoming increasingly apparent (Lau et al., 2013;Satoh et al., 2009).
In contrast, ECM remodeling benefits remyelination upon toxininduced demyelination in rodent models, pointing to abnormal ECM remodeling in MS lesions, which is reviewed next.

| THE INTER S TITIAL ECM UP ON DEMYELINATING INJ URY
Dynamic remodeling of the ECM, that is transient expression and/ or degradation, is an effective mechanism to regulate glial cell behavior, including OPCs upon injury (reviewed in (Pu, Stephenson, & Yong, 2018)). Analysis of mRNA expression of ECM proteins and the MS lesion proteome reveals that mixed active/inactive and inactive MS lesions have a unique ECM composition, both when compared among each other and to control white matter and active MS lesions (Hendrickx et al., 2017;Mohan et al., 2010;Satoh et al., 2009). This hints to interlesional heterogeneity of ECM proteins in MS lesions and that dysregulated ECM remodeling may play an important role in the pathogenesis of MS, including remyelination failure. To unravel abnormal ECM composition in the distinct MS lesions, experimental toxin-induced demyelination rodent models that show robust remyelination in the absence of the complex inflammatory background in MS are most instructive to obtain insight into the natural cellular and molecular responses during demyelination of the CNS and subsequent successful recovery. These toxin-induced demyelination models include the dietary cuprizone model (mice), which shows global demyelination most prominently in the corpus callosum (Praet, Guglielmetti, Berneman, Van der Linden, & Ponsaerts, 2014;Skripuletz, Gudi, Hackstette, & Stangel, 2011;Torkildsen, Brunborg, Myhr, & Bo, 2008), and the focal lysolecithin (mice and rat) and ethidium bromide (mice) models, where demyelination is induced by local injection of the toxin in the area of interest (Blakemore, 1976;Jeffery & Blakemore, 1995;. Notably, in the lysolecithin model the BBB is disrupted (Muramatsu et al., 2015), while the endothelial cells remain seemingly intact in the cuprizone and ethidium bromide model (Bjelobaba, Begovic-Kupresanin, Pekovic, & Lavrnja, 2018;Kondo, Nakano, & Suzuki, 1987;McMahon, Suzuki, & Matsushima, 2002). In the following, we outline for each ECM protein transient changes in its expression in the interstitial matrix, that is parenchymal ECM, upon white matter demyelination and during remyelination in experimental rodent models, and indicate its presence in the distinct white matter MS lesions (summarized in Table 1 and Figure 1). In addition, we discuss direct effect(s) of each ECM protein on OPC behavior (summarized in Figure 2) and discuss how the protein may contribute to remyelination (failure). We only briefly touch upon changes in ECM composition of basement membranes, as remyelination-predetermined OPCs in general do not face the ECM in basement membranes. For irregularities and alterations of ECM in basement membranes, we refer to two excellent reviews by van Horssen et al. (2007) and Lau et al. (2013).

| Structural ECM proteins
CSPGs are the main proteoglycans of the CNS and consist of core proteins with covalently linked sulfated chrondroitin GAG side chains. The individual secreted CSPGs, that is neurocan, aggrecan, versican, and phosphacan, differ in the composition of their core protein and/or the number and type of attached GAG chains.
Note: Increased (↑), decreased (↓), and similar (=) expression of interstitial ECM proteins upon toxin-induced demyelination and in distinct white matter MS lesions compared to control.
Depositions of CSPGs, including aggrecan, neurocan and versican, are present in astrocyte-enriched edges of active and inactive MS lesions, while their expression is downregulated in the center of active, mixed active/inactive and in inactive lesions (Dahl, Perides, & Bignami, 1989;Hendrickx et al., 2017;Sobel & Ahmed, 2001).
Gene expression profiling of rim and center of mixed active/ inactive and inactive MS lesions revealed that neurocan mRNA is enriched at the rim of mixed active/inactive lesions (Hendrickx et al., 2017). Deposited CSPG components in active MS lesions are suggested to be phagocytosed by foamy macrophages together with myelin debris (Sobel & Ahmed, 2001), although expression of CSPGs by macrophages cannot be excluded (Lau et al., 2012).
Strikingly, phosphacan expression, which is a CNS-specific CSPG, is more or less preserved in active lesions, and, in general, less reduced in inactive MS lesions compared to the other CSPGs (Sobel & Ahmed, 2001). Also, in situ hybridization studies show the presence of phosphacan mRNA in remyelinating oligodendrocytes in MS lesions (Harroch et al., 2002), and suggested to be involved in oligodendrocyte lineage progression (Karus et al., 2016). Of interest, granular aggregates of versican and aggrecan, and to a lesser extent of neurocan, but not phosphacan are evident in normal appearing white matter (NAWM; Sobel & Ahmed, 2001).
Thus, demyelination in experimental models leads to the upregulation of CSPGs, which may have beneficial functions at early stages F I G U R E 2 Effect of extracellular matrix (ECM) proteins on oligodendrocyte progenitor cell (OPC) behavior relevant to remyelination. Successful remyelination upon central nervous system (CNS) demyelination in experimental rodent models involves the activation of adjacent OPCs, followed by recruitment (migration and proliferation) to and maturation (process arborization, differentiation, and myelin membrane formation) in the demyelinated area. The effect of structural ECM proteins, fibrous ECM glycoproteins, and matricellular proteins that are present in the interstitial ECM upon toxin-induced CNS white matter demyelination and/or distinct white matter MS lesions are shown. ECM proteins depicted in green increase and ECM proteins depicted in red decrease the indicated OPC behavior [Color figure can be viewed at wileyonlinelibrary.com] of recovery, that is to prevent premature OPC differentiation.
Moreover, CSPGs are cleared to enable remyelination. Deposition of CSPGs in MS lesion edges may lead to the formation of a barrier for OPC migration into the lesions, and loss of CSPGs in the center may preclude their beneficial actions after recovery (Keough et al., 2016;Lau et al., 2012;Pu et al., 2018;Sobel & Ahmed, 2001).
Hyaluronan is a specialized non-sulfated GAG that functions as free, that is independent GAG molecule without protein core, or engages in non-covalent interactions with proteoglycans, including CSPGs (Sherman et al., 2002). Hyaluronan is expressed in different sizes, performing its functions in cell growth and motility and interactions between different ECM molecules to stabilize the ECM (Bignami et al., 1993). Upon lysolecithin-induced demyelination, only a minimal accumulation of hyaluronan is noticed (Back et al., 2005), suggesting that there is no major upregulation of hyaluronan after demyelinating injury in this experimental rodent model. However, deposition of hyaluronan is associated with MS development (Nagy et al., 2019). In contrast to other glycosaminoglycans (Sobel & Ahmed, 2001), hyaluronan accumulates in the core of inflammatory demyelinating active MS lesions (Back et al., 2005). In early lesions, infiltrating T cells and microglia probably produce hyaluronan and in chronic lesions, astrocytes are the likely source (Back et al., 2005). Also, OPCs, and to a lesser extent mature oligodendrocytes, synthesize hyaluronan (Preston et al., 2013). Most interestingly, T cells and microglia produce low-molecular weight hyaluronan ranging from 200 to 400 kDa (LMW hyaluronan), whereas astrocytes produce hyaluronan with high-molecular weight, ranging from 900 to 1,000 kDa (HMW hyaluronan; Back et al., 2005). Both HMW and LMW hyaluronan are present in MS lesions (Back et al., 2005;Preston et al., 2013). HMW hyaluronan is detrimental for remyelination when injected in lysolecithin-induced demyelinated lesions (Back et al., 2005). Treatment with soluble HMW hyaluronan or LMW hyaluronan inhibits OPC maturation through LMW hyaluronanmediated activation of TLR2 (Back et al., 2005;Sloane et al., 2010).
In fact, OPC maturation is only precluded when HMW hyaluronan is processed to a LMW form by a specific hyaluronidase (PH20) that is present in OPCs and astrocytes in lysolecithin-induced lesions and MS lesions (Back et al., 2005;Preston et al., 2013). Hence, digestion products of HMW hyaluronan as well as unprocessed LMW hyaluronan likely prevent OPC maturation in chronic MS lesions, thereby contributing to remyelination failure ( Figure 2; Preston et al., 2013).
Collagens are trimeric proteins containing long triple helical sequences that have the ability to form stable fibrils. They are arranged into networks and are involved in structuring and providing rigidity to the ECM. In the adult CNS, collagen is mainly limited to the basement membranes (e.g., collagen IV) and is hardly present in the interstitial matrix, which makes the brain a relatively soft tissue.
While fibrillar collagens are not present in CNS parenchyma upon cuprizone-induced demyelination (Hibbits et al., 2012), fibrillar collagen V is closely associated with astrocytes in the interstitial matrix of active MS lesions (Mohan et al., 2010). Furthermore, genes involved in collagen synthesis are highly expressed in the rim of inactive MS lesions (Hendrickx et al., 2017). Also, the accumulation of fibrillary collagens I and II is associated with perivascular inflammation in the center of active and mixed active/inactive MS lesions, and mainly restricted to basement membranes, playing a role in limiting the enlargement of MS lesions (Mohan et al., 2010). MS proteome analyses demonstrated an enrichment of collagen IV, primarily localizing in the basement membranes in inactive lesions (Satoh et al., 2009).
Although OPCs lack collagen-recognizing integrins, OPC migration is inhibited on collagen I substrates (Milner, Edwards, Streuli, & ffrench-Constant, 1996), and collagen I microspheres support OPC differentiation (Yao, Phan, & Li, 2013). Therefore, the contribution of interstitial collagen to remyelination failure via direct modulation of OPC behavior in MS lesions is likely negligible.
Next to localized synthesis, part of the vitronectin acquires access into the brain, following passage across the disrupted BBB (Sobel et al., 1995). Vitronectin is absent in inactive lesions (Sobel et al., 1995). It promotes migration and proliferation of cultured OPCs (Baron, Shattil, & ffrench-Constant, 2002;Frost, Kiernan, Faissner, & ffrench-Constant, 1996), indicating that vitronectin may regulate OPC recruitment, and that the absence of the compound in chronic lesions may contribute to the reduced number of OPCs in white matter MS lesions.
Laminins are self-polymerized and heterotrimeric glycoproteins that serve as major adhesive proteins in basement membranes.

| Matricellular proteins
Matricellular proteins are non-structural regulators of the ECM that contain binding sites for other ECM proteins as well as cell surface receptors, and therefore play important roles in controlling cell behavior and ECM remodeling (Bornstein & Sage, 2002).
Thrombospondin-1 is a trimeric matricellular protein that is present in the adult CNS (Asch, Leung, Shapiro, & Nachman, 1986) Osteopontin is a secreted matricellular protein that binds directly to fibronectin and collagen (Giachelli & Steitz, 2000), and is expressed in grey but not in white matter in the normal adult CNS (Selvaraju et al., 2004;Shin, Cha, Chun, Chung, & Lee, 1999;Zhao, Fancy, ffrench-Constant, & Franklin, 2008). However, the protein is transiently upregulated in white matter upon cuprizone- (Selvaraju et al., 2004) and ethidium bromide-induced (Zhao, Fancy, ffrench-Constant, & Franklin, 2008) demyelination. Microglia/ macrophages and astrocytes show osteopontin immune reactivity (Selvaraju et al., 2004;Zhao et al., 2008), while only microglia/ macrophages harbor osteopontin mRNA (Zhao et al., 2008), and likely secrete osteopontin in the demyelinated areas. Microarray analyses showed that osteopontin mRNA levels are higher in active MS lesions than in control white matter (Chabas et al., 2001). In addition, proteomic analyses revealed that osteopontin protein is is also present in astrocytes in NAWM at areas with high levels of microglia activation (Chabas et al., 2001;Sinclair et al., 2005) and occasionally in white matter oligodendrocytes (Chabas et al., 2001;Diaz-Sanchez et al., 2006). Contrastingly, while protein levels are increased, gene expression analysis of adjacent NAWM, rim and center of mixed active/inactive and inactive lesions revealed a downregulation in osteopontin transcript levels (Koning, Bo, Hoek, & Huitinga, 2007). In vitro studies demonstrated that soluble osteopontin induces proliferation in OPC-like cell lines and increases the expression of myelin basic protein (MBP) and myelin membrane formation in mixed cortical cultures (Selvaraju et al., 2004). Although proliferation and differentiation are seemingly opposite processes, both are necessary for successful remyelination. Similar to an integrin-and developmental stage-dependent switch in growth factor signaling (Baron, Colognato, & ffrench-Constant, 2005;Baron et al., 2002;Colognato et al., 2002), the effect of osteopontin on OPC behavior may depend on the developmental stage and to which integrin it binds. Thus, osteopontin may mediate OPC proliferation via integrin αvβ3, whereas OPC differentiation may be induced by signaling via integrin αvβ5 (Blaschuk, Frost, & ffrench-Constant, 2000). In favor of a developmental stage-dependent effect is that osteopontin is added to differentiating mixed cortical cultures (Selvaraju et al., 2004), that is when OPC proliferation is ceased. Alternatively

| Role of abnormal ECM in remyelination failure
Similar to demyelinating CNS white matter injury in experimental models, ECM molecules that are absent in the healthy adult CNS are upregulated upon demyelination in MS lesions, indicating a "normal" initial response to demyelination in MS lesions. These interstitial ECM proteins in general contribute to OPC migration and proliferation ( Figure 2), and prevent premature OPC differentiation. In experimental models where remyelination is successful, the proteins are transiently expressed and cleared at the onset of remyelination to allow for OPC differentiation, among others by the increased expression of laminin (Zhao et al., 2008). In MS lesions, the initial transient increased ECM proteins persist, while the remyelination beneficial ECM protein laminin is virtually absent in the interstitial ECM MS. This indicates that the "normal" ECM remodeling response in MS lesions is derailed and not converted to a "remyelination-favoring mode" (Table 1 and Figure 1).
Another striking difference is that CSPGs, present in the interstitial  (Table 1 and Figure 1). Indeed, remyelination at the rim is more pronounced than that at the MS lesion center (Raine & Wu, 1993), suggesting that the interstitial ECM at the edges of active and mixed active/inactive lesions is more permissive for remyelination than the ECM in the center of the lesions. However, the degree of remyelination depends on the net ratio of myelination-supportive (e.g., laminin) and myelination-inhibitory ECM molecules (e.g., fibronectin aggregates, hyaluronan, CSPGs

| Role of glial scar and inflammation
Astrocytes, microglia, and macrophages are the producers of ECM proteins in MS lesions. The astrocyte response in MS lesions is of dual nature, that is both astrocyte loss and astrogliosis are associated with remyelination failure (Correale & Farez, 2015;Nair, Frederick, & Miller, 2008;Williams et al., 2007). Major barriers that contribute to remyelination failure in MS are glial scars, that is astrogliosis, which parallels ECM alterations, and the ensuing inflammation. The changes in reactive astrogliosis are regulated in a context-specific manner (Sofroniew & Vinters, 2010), and given that the distinct   (Stoffels et al., 2013) accumulate in lesioned areas. Also, CSPG proteins neurocan, aggrecan, brevican, and versican V1 are also increased at the peak of clinical EAE severity, whereas the expression of versican V2 is reduced (Sajad, Zargan, Chawla, Umar, & Khan, 2011;Stephenson et al., 2018). Aggrecan is only present in spinal cord grey matter, while versican V1 is also present in perivascular cuffs and closely associated with infiltrating immune cells (Sobel & Ahmed, 2001). In TME, which is characterized by mild inflammation and insufficient OPC differentiation ( (Miron et al., 2013;Peferoen et al., 2015;Vogel et al., 2013), the persistent or incorrect expression of ECM molecules in MS lesions might result from altered expression patterns of microglia/macrophage-derived MMPs (Kieseier et al., 1999), which is discussed next.

| MMPs UP ON DEMYELINATING INJ URY
MMPs are a family of proteolytic enzymes, also referred to as endopeptidases, that are essential for ECM remodeling in many processes. These include migration, wound healing, tissue morphogenesis, cell differentiation, neuronal growth, and several signaling processes (Lu et al., 2011;Page-McCaw et al., 2007). A propeptide region is located at the N terminus of the protein, which prevents proteolytic activity through interaction with the catalytic domain. This domain is removed to activate the enzyme.
The catalytic domain is also located at the N terminus, in which a  Lu et al., 2011;Page-McCaw et al., 2007).
In the healthy adult CNS, MMP activity is crucial in supporting cognitive processes, such as learning and memory, due to ECM remodeling and the regulation of synaptic plasticity and long-term potentiation (Agrawal, Lau, & Yong, 2008;Huntley, 2012). Also, the physiology of axons, myelin turnover, and angiogenesis are regulated by several MMPs. In addition, differential expression of MMPs in the development of the CNS is essential for neurogenesis and axonal growth as well as for the function of oligodendrocytes and myelinogenesis. MMPs also play important roles in repair processes and in pathology, and both beneficial and detrimental functions have been assigned to MMPs in the injured CNS (Javaid, Abdallah, Ahmed, & Sheikh, 2013;Yong, 2005;Yong, Power, Forsyth, & Edwards, 2001).
To However, uncontrolled and abundant expression of MMPs may damage the BBB, induce inflammation, and neurotoxicity, which may lead to (demyelinating) injury (Agrawal et al., 2008;Yong, 2005;Yong et al., 2001). Of interest, synergism among MMP2, MMP9, and MMP7 genes may be a susceptibility factor for MS (Rahimi et al., 2016). In the following, we provide an overview of current insight into the role of distinct MMPs in successful remyelination and their expression in the distinct MS white matter lesions (summarized in Table 2 and Figure 1).
Also, the consequence for demyelinating pathology is discussed. Škuljec et al. (2011)  demyelination and subsequent remyelination. At early demyelination, MMP3 mRNA is upregulated, followed by a return to normal levels during late stages of demyelination. A second more prominent upregulation of MMP3 mRNA is observed upon remyelination, which is also reflected at the protein level (Skuljec et al., 2011;Wang et al., 2018). MMP3 mRNA levels, but not protein levels, are also increased upon early remyelination in the lysolecithin-induced demyelination model . At the protein level, MMP3, also referred to as stromelysin-1, is predominantly expressed by astrocytes, but may also be produced by damaged neurons, microglia, and oligodendrocytes following other types of CNS injury (Van Hove, Lemmens, Van de Velde, Verslegers, & Moons, 2012). An (early) upregulation of MMP3 is also observed in the inflammatory TME model (Hansmann et al., 2012;Ulrich et al., 2006), and in EAE (Weaver et al., 2005). In these inflammatory models with insufficient remyelination, MMP3 may aid to the disruption of the BBB. In addition, the early upregulation of MMP3 may cause the breakdown of myelin and exacerbate demyelination (Chandler, Cossins, Lury, & Wells, 1996;Shiryaev et al., 2009)  . However, MMP3 protein is present in hypertrophic astrocytes in active and chronic white matter MS lesions, in microglia/macrophages in active lesions and on vasculature in active and mixed active/inactive MS lesions (Maeda & Sobel, 1996;Wang et al., 2018). Biochemical analysis revealed that MMP3 protein levels were increased in mixed active/inactive MS lesions, but not inactive lesions compared to control white matter . Of interest, MMP3 is a potent activator of other MMPs, such as MMP7 and MMP9 (Lu et al., 2011;Maeda & Sobel, 1996;Van Hove et al., 2012), which are also prominently present in MS lesions (see below).

| MMP12
The expression level of MMP12, or macrophage elastase, is significantly altered upon cuprizone-induced demyelination. At demyelination, MMP12 mRNA is predominantly produced by microglia/ macrophages (Skuljec et al., 2011). In this phase, MMP12 may function in cellular migration of macrophages through ECM remodeling and degeneration of myelin membranes through cleavage of MBP, one of the major myelin components (Chandler et al., 1996;Gronski et al., 1997;Shipley, Wesselschmidt, Kobayashi, Ley, & Shapiro, 1996). The elevated expression pattern continued during remyelination, where MMP12 is produced by astrocytes and to some extent by oligodendrocytes, rather than microglia/macrophages (Skuljec et al., 2011). Upon remyelination, astrocyte-derived MMP12 may contribute to clearance of both myelin debris and the transientexpressed ECM proteins, and thus induce a stimulatory environment for remyelination. Also, during CNS development, MMP12 releases IGF-1 from IGF binding protein 6 and is required for process elongation and OPC differentiation (Larsen, DaSilva, Conant, & Yong, 2006;Larsen & Yong, 2004). Studies in EAE and TME models demonstrated an upregulation of MMP12 by microglia/macrophages and a suggested role in disease progression and chronic phases of demyelination, respectively (Dasilva & Yong, 2008;Hansmann et al., 2012;Ulrich et al., 2006). In contrast, the EAE disease course is worse in MMP12 knockout than in wild type mice, which is mediated in part by modulating the Th1/Th2 effector cytokine balance (Weaver et al., 2005) and MMP12-mediated cleavage of osteopontin (Goncalves DaSilva, Liaw, & Yong, 2010). In fact, osteopontin is linked to relapses by enhancing the survival of activated T cells (Hur et al., 2007). This indicates that MMP12 is a protective molecule in EAE. In TME, MMP12 is likely involved in demyelination and extravasation of macrophages, and not in BBB damage or ECM remodeling (Hansmann et al., 2012). In MS lesions, MMP12 protein is present in foamy macrophages and upregulated within active demyelinating lesions, and at the rim of inactive MS lesions and to a lesser extent in the center of mixed active/inactive and inactive lesions (Vos, van Haastert, de Groot, van der Valk, & de Vries, 2003). In contrast to cuprizone-induced demyelination, in MS lesions, MMP12 is not observed in astrocytes and oligodendrocytes. Hence, the pre-

| MMP9
MMP9, also called gelatinase B, is one of the most studied MMPs in MS. While MMP9 mRNA expression is unaltered upon cuprizoneinduced demyelination compared to unlesioned white matter tissue (Skuljec et al., 2011), MMP9 protein levels are upregulated predominantly in microglia and macrophages at the onset of remyelination upon lysolecithin-induced demyelination (Larsen, Wells, Stallcup, Opdenakker, & Yong, 2003). MMP9 mRNA and protein are upregulated in active and mixed active/inactive demyelinating MS lesions (Anthony et al., 1997;Cossins et al., 1997;Cuzner et al., 1996;Lindberg et al., 2001;Maeda & Sobel, 1996;Mohan et al., 2010). MMP9 mRNA expression is also upregulated in an EAE model Kieseier et al., 1998;Weaver et al., 2005), but not in the TME model (Ulrich et al., 2006(Ulrich et al., , 2008. It is hypothesized that MMP9 is important for the infiltration of inflammatory cells into the CNS. Thus, the level of MMP9 is significantly increased in CSF and serum of MS patients with active disease, compared to healthy individuals (Bar-Or et al., 2003;Benesova et al., 2009;Fainardi et al., 2006;Waubant et al., 1999). Injection of recombinant MMP9 into the CNS parenchyma resulted in breakdown of the BBB, neuronal and myelin loss (Anthony et al., 1998). The localization of MMP9 in cells present in perivascular areas, such as endothelial cells and the infiltrating cell population, including lymphocytes and macrophages, is also in favor of its contribution to the disruption of the BBB in MS Lindberg et al., 2001). In addition, resident CNS cells, such as microglia/macrophages and astrocytes in active MS lesions also express MMP9, mainly around the perivascular areas but also in the CNS parenchyma (Cuzner et al., 1996;Maeda & Sobel, 1996). In mixed active/inactive lesions, MMP9 is more prominently localized at the edge of the lesions (Anthony et al., 1997), while this protease is occasionally also present in astrocytes in inactive lesions. These data further suggest that MMP9 is upregulated and exert its functions in active demyelinating lesions during inflammation. In addition to the stimulation of inflammatory cell infiltration through BBB disruptions, MMP9 derived from leukocytes cleaves MBP and may be involved in the degradation of myelin membranes (Gijbels et al., 1993;Proost, Van Damme, & Opdenakker, 1993). In contrast, MMP9 has also beneficial roles for remyelination (Larsen et al., 2003;Oh et al., 1999;Siskova et al., 2009;Uhm, Dooley, Oh, & Yong, 1998). Localization of MMP9 at the tips of the extending processes of oligodendrocytes is essential for their outgrowth, that is process elongation and arborization is reduced in absence of MMP9 (Oh et al., 1999;Uhm et al., 1998) and upon mislocalization of its activity (Siskova et al., 2009). In addition, MMP9 produced by macrophages and microglia in the remyelination phase degrades NG2, a membrane-spanning CSPG present on OPCs that inhibits maturation of oligodendrocytes (Larsen et al., 2003). Hence, MMP9 plays a dual role in MS. The activity of this enzyme in active demyelinating lesions may be detrimental and contribute to the pathogenesis and progression of MS, but beneficial functions of MMP9 are observed for remyelination, in which it facilitates oligodendrocyte maturation via ECM remodeling.

| MMP2
MMP2, is another member of the gelatinase family, also referred to as gelatinase A, and is in contrast to MMP3, MMP9, and MMP12, constitutively expressed in the CNS and CSF (Anthony et al., 1997;Rosenberg, 2002). MMP2 mRNA expression is not altered during demyelination and remyelination in a cuprizone-induced demyelination model (Skuljec et al., 2011), and only minor upregulation is observed in EAE (Weaver et al., 2005), in TME (Ulrich et al., 2006) and active MS lesions Mohan et al., 2010). However, MMP2 protein is present in macrophages and infiltrating cells in the perivascular area of active MS lesions (Anthony et al., 1997;Diaz-Sanchez et al., 2006;Maeda & Sobel, 1996). MMP2-expressing macrophages are also present at the border of mixed active/inactive lesions, while its expression in chronic lesions is only perivascular (Anthony et al., 1997;Diaz-Sanchez et al., 2006). As MMP9, MMP2 is likely involved in disruption of the BBB (Rosenberg et al., 1992) and has the highest activity in MBP degradation (Chandler et al., 1995;Diaz-Sanchez et al., 2006). MMP2 appears to be more abundant in MS lesions than MMP9, and its expression is predominant in areas of damaged axons, particular at lesion borders (Diaz-Sanchez et al., 2006). Interestingly, MMP2 is also upregulated in NAWM areas adjacent to the lesions (Anthony et al., 1997;Diaz-Sanchez et al., 2006;Maeda & Sobel, 1996). Also, higher levels of MMP2 are present in serum of MS patients (Bar-Or et al., 2003;Benesova et al., 2009).
Whether MMP2 may play a role in regeneration of myelin other than its potential to locally degrade only CSPGs, but not laminin, which is present in the same area (Zuo, Ferguson, Hernandez, Stetler-Stevenson, & Muir, 1998), remains to be determined.

| MMP7
MMP7, also called matrilysin, is constitutively expressed in the brain (Anthony et al., 1997;Wang et al., 2018). However, its mRNA levels are at the lower limit of detection Mohan et al., 2010). Given that MMP7 has a potent activity and a broad substrate specificity, it has been suggested that MMP7 may be a regulator of ECM turnover in the healthy brain (Anthony et al., 1997). MMP7 mRNA expression is neither increased upon cuprizone-induced demyelination (Skuljec et al., 2011) nor in the TME (Ulrich et al., 2006) model, while MMP7 mRNA levels were increased upon lysolecithininduced demyelination, and early remyelination .
In the latter model, MMP7 is localized extracellularly and present in microglia/macrophages . Contrasting findings are reported in EAE models. In a mouse MOG peptide-induced EAE model MMP7 mRNA levels remain similar (Weaver et al., 2005), while MMP7 mRNA is increased by 500-fold during the course and as protein present in invading macrophages in a MOG-induced rat EAE model  and increased at the peak of an adoptive-transfer rat EAE model (Kieseier et al., 1998). Also, contrasting findings have been reported for MMP7 mRNA levels in MS lesions: MMP7 mRNA expression is increased in all lesion types in some studies Lindberg et al., 2001), while it was undetectable in active and chronic MS lesions in another study (Mohan et al., 2010). MMP7 protein is localized to parenchymal macrophages and occasionally observed in astrocytes in active MS lesions with a weaker expression in the center than at the edge of the lesion Wang et al., 2018). MMP7 expression in macrophages is also prominent at the lesion borders of mixed active/ inactive MS lesions (Anthony et al., 1997) and not as prominent in the center , while biochemical analysis reveals that proMMP7 expression is reduced in mixed active/inactive and inactive lesions . In remyelinated lesions, total expression levels of MMP7 are comparable to control white matter of healthy subjects with occasional expression in macrophages . MMP7 is not elevated in serum of MS patients, which is in contrast to MMP9 and MMP2 (Bar-Or et al., 2003). The exact role of MMP7 is not well understood, but like the other MMPs, a role in extravasation of monocytes into the tissue and migration of macrophages through remodeling of basement membrane ECM, such as proteoglycans, fibronectin, laminin, and elastin, can be foreseen. In addition, MMP7 potentially induces bystander demyelination and axonal loss, although likely to a lesser extent than MMP2 and MMP9 (Anthony et al., 1997). However, when expressed in the CNS parenchyma in a timely manner, MMP7 may also be beneficial in clearing the transiently expressed ECM components and myelin debris (Chandler et al., 1995).

| Other MMPs
Other MMPs that have been analyzed at the protein level in MS lesions are MMP1 (Maeda & Sobel, 1996), MMP19 (van Horssen et al., 2006), and MMP28 (Werner, Dotzlaf, & Smith, 2008). MMP1 is expressed in the majority of macrophages in active MS lesions (Maeda & Sobel, 1996). MMP19 is constitutively expressed by microglia in the healthy adult CNS, and highly expressed in macrophages in the parenchyma and perivascular areas in active MS lesions and in the rim of mixed active/inactive lesions, and occasionally by reactive astrocytes (van Horssen et al., 2006). Also, MMP19 mRNA levels are increased in active and inactive MS lesions (Mohan et al., 2010). While the function of MMP19 in MS pathology remains to be established, it is interesting to note that MMP19 specifically degrades the large isoform of tenascin-C of which the expression is reduced in active MS lesions and mixed active/inactive lesions borders (Table 1 and Figure 1; Stracke et al., 2000). Upregulated MMP28 expression has been shown in one demyelinated, uncharacterized MS lesion, and in EAE (Werner et al., 2008). In addition, MMP28 mRNA is increased in inactive MS lesions (Mohan et al., 2010). Moreover, Western blot analysis also shows a marked upregulation in NAWM compared to control white matter (Werner et al., 2008). In the adult CNS, MMP28 is mainly expressed in neurons and is a negative regulator of myelination (Werner et al., 2008) and macrophage recruitment (Manicone et al., 2009), justifying more research on this protein. lesions, respectively (Mohan et al., 2010). The cellular localization of these MMPs and their significance for ECM remodeling, remyelination, and MS lesion pathology remains to be determined. Of note, the expression levels of most of these MMPs are also altered in TME and EAE models (Ulrich et al., 2006;Weaver et al., 2005).

| MMP DYS FUN C TI ON IN MS LE S I ON S
MMPs benefit remyelination by clearing the transiently expressed ECM proteins upon demyelination (Figure 4a). In MS lesions, ECM substrates are not appropriately degraded by MMPs, among others by the absence of MMPs in the lesion center and the inability of MMPs to process substrates at the lesion rim ( Figure 4b). In addition, the interference of MMPs with basement membranes contributes to demyelination. In the following, we reflect on MMP dysfunction and how this may contribute to abnormal ECM expression in MS lesions.

| Altered MMP expression
The laminin, which enables the inflammatory cells, that is lymphocytes and monocytes to infiltrate the CNS (Leppert, Lindberg, Kappos, & Leib, 2001). Also, most MMPs present in inflammatory lesions degrade MBP, which may enhance demyelination and, in turn, contributes to axonal damage (Anthony et al., 1998). However, since MBP is an intracellular, peripheral protein, prior myelin degeneration is obviously required to allow proteolytic accessibility of the protein. In fact, the potential of MMPs to degrade MBP may indirectly facilitate clearance of remyelination-inhibiting myelin debris (Kotter et al., 2006).  Tables 1 and 2). The tightly and temporal (cellular) expression of several MMPs regulate the clearance of the transiently expressed ECM molecules to enable regeneration of myelin (Figure 4a). In contrast, in MS lesions, where remyelination fails, a different and more persistent pattern of MMPs and ECM molecules is observed (Figure 4b; Agrawal et al., 2008;van Horssen et al., 2007). For example, the mRNA levels of gelatinases MMP2 and MMP9 are not (MMP2) or transiently (MMP9) upregulated upon demyelination and remyelination models (Larsen et al., 2003;Skuljec et al., 2011), but these enzymes are highly expressed in active and mixed active/inactive MS lesions (Anthony et al., 1997;Cossins et al., 1997;Cuzner et al., 1996;Lindberg et al., 2001;Maeda & Sobel, 1996). The disturbed ECM in MS lesions may regulate the expression of these MMPs.

| MMPs are present at the rim of chronic MS lesions
MMP7 and MMP3 are more potent proteases to degrade the interstitial ECM, present in MS lesions (Imai et al., 1995;Muir et al., 2002;Murphy et al., 1991;Siri et al., 1995). Therefore, these MMPs are potentially able to reintroduce a permissive environment for remyelination and improve the neuropathological conditions in MS. Indeed, both MMPs are upregulated at remyelination following demyelination in experimental rodent models (Skuljec et al., 2011;Wang et al., 2018). MMP3 degrades the CSPGs, aggrecan, versican, neurocan, and phosphacan, while MMP7 cleaves aggrecan (Fosang et al., 1992;Muir et al., 2002 (Cuzner et al., 1996). Another tightly regulated mechanism for MMP actions is the regulation of its activity by TIMPs (Lu et al., 2011;Page-McCaw et al., 2007). Upon toxin-induced demyelination, increased mRNA levels of TIMP1 and TIMP2 are observed at demyelination, followed by a gradual decline during remyelination (Skuljec et al., 2011;Wang et al., 2018). TIMP3 is transiently upregulated in the first week during demyelination, whereas TIMP4 is upregulated during remyelination (Skuljec et al., 2011). Lindberg et al. (2001) show that the mRNA levels of these TIMPS are not significantly altered at all MS lesion types. In contrast, Mohan et al. (2010) show that both TIMP1 and TIMP3 mRNA are upregulated in active and inactive lesions, while TIMP3 mRNA are downregulated in inactive lesions. Immunohistochemistry of TIMPs may reveal the expression patterns and cellular localization of TIMPs, and whether the delicate balance between MMP and their inhibitors is disturbed in MS lesions, that is interfere with interstitial ECM remodeling. Also, an in situ MMP activity assay on MS lesions may provide insight in localized MMP activity levels. Of note, inhibiting MMP activity reduced the clinical symptoms in an EAE model by reducing the breakdown of the BBB and demyelinating pathology (Gijbels, Galardy, & Steinman, 1994;Liedtke et al., 1998). Hence, although modulation of MMP activity can be an advantageous approach, it remains to be determined whether MMP activity within the CNS is actually inhibited in this model.
In contrast to CSPGs, tenascin expression is reduced in active lesions and the lesion border of inactive lesions, indicating proteolytic activity at these sites. Upon alternative splicing, a large and small isoform of tenascin-C are generated. The small variant is more resistant to degradation, but cleaved by MMP7, while MMP3, MMP7, MMP12, and MMP19 degrade large tenascin-C. Also, other, yet undefined MMPs, or even non-related proteases, such as cathepsin B (Bever & Garver, 1995;Mai, Sameni, Mikkelsen, & Sloane, 2002), may clear tenascins in active MS lesions and at the border of mixed active/inactive lesions.

| MMPs are absent in the center of chronic MS lesion
MMPs are hardly present in inactive and in the center of mixed active/inactive lesions, which may be the main reason why the ECM proteins, fibronectin and osteopontin, persist. While loss of function studies demonstrated that fibronectin (Stoffels et al., 2015) and osteopontin (Zhao et al., 2008) are redundant for remyelination, the persistence of otherwise transient ECM proteins in MS lesions, such as fibronectin (aggregates), results in a gain of function, that is they perturb OPC differentiation. Why MMPs are not upregulated in chronic MS lesions is not known. It may be hypothesized that the cellular source of MMPs might determine whether these enzymes perform beneficial or detrimental roles in MS lesions. This is supported by observations that MMP12 is initially produced by microglia/macrophages, while the protease localizes to astrocytes and oligodendrocytes, as observed at successful remyelination, which is beneficial for remyelination. Also, the CSPG-enriched barrier at the lesion border may prevent the migration of cells that produce a suitable MMP to the lesion center. CSPGs are reduced in the lesion center, indicating that they are cleared, but not resynthesized to reestablish the interstitial ECM in adult healthy CNS. Indeed, mRNA levels of CSPGs are reduced in inactive MS lesions (Mohan et al., 2010), which may be due to the lack of sufficient and/or the presence of misactivated microglia/macrophages, the main producers of CSPGs, in the center of lysolecithin-induced lesions (Lau et al., 2012). In this regard, it will be of interest to analyze the ECM composition of remyelinated lesions. In remyelinated lesions, MMP7 levels are comparable to control white matter , while fibronectin levels are still increased (Stoffels et al., 2013).

| CON CLUD ING REMARK S AND PER S PEC TIVE S
Upon demyelinating injury, remodeling of the interstitial ECM is essential in guiding oligodendrocyte behavior to succeed successful remyelination. Remodeling of the interstitial ECM upon demyelination in MS lesions, including the expression patterns of ECM molecules and MMPs, is dependent on the nature of and the localization within lesions, and differs from ECM remodeling upon demyelinating injury in experimental rodent models that show successful recovery (Figure 4). More specifically, the transient upregulation of ECM proteins, as a natural response to demyelination, persists in MS lesions, among others by dysregulation of MMP expression, their localization and, likely, their activation.
Several MMPs are upregulated in a seemingly uncontrollable manner (MMP2 and MMP9), while others (MMP3, MMP7, and MMP12) that perform beneficial functions for remyelination upon demyelination in CNS white matter in rodent models are absent or incorrectly located in MS lesions. Also, and in contrast to demyelination in experimental models, MMPs are mainly expressed by microglia/ macrophages and hardly, with the exception of MMP3, by astrocytes ( Figure 4). In addition, the phenotype of astrocytes, producers of ECM molecules in the CNS, is changed as a consequence of the ensuing spatially focused inflammation, leading to the formation of at least two glial scars: one at the border of active and inactive lesions, which is enriched in CSPGs, and the other within the center of MS lesions that is devoid of CSPGs, but enriched in fibronectin aggregates, osteopontin, and hyaluronan.
Taking the effect on OPC behavior into account, the persistent presence of mainly OPC differentiation-inhibiting ECM proteins and the absence of myelination-promoting laminin in MS lesions contribute to remyelination failure. As a therapeutic strategy, it is essential to clear or bypass these dominant OPC differentiation-inhibiting ECM proteins, as targeted upregulation of remyelination-promoting laminin may not be efficient to enable regeneration of myelin.
Therefore, the ECM environment in MS lesions should also be taken into account when determining the effectiveness of potential remyelination-inducing compounds that are based on stimulating endogenous OPCs or when a cell therapy with exogenous remyelinating cells is considered as a strategy to promote remyelination in MS lesions.
Local activation of MMPs, that is in the rim and/or lesions center with an appropriate substrate specificity, may be a means to degrade ECM depositions and reinstate a permissive environment for remyelination, as is the case in recovery upon demyelination in experimental models. However, the dual role of MMPs, such as their ability to degrade vascular basement membrane ECM components and to induce axonal injury and myelin loss, complicates the development of treatment strategies, aimed at correct interstitial ECM remodeling in favor of remyelination. Thus, MMPs contribute to damage at early lesions stages, while their absence at later stages is detrimental to interstitial ECM remodeling, which, in turn, is essential for the regeneration of myelin. In fact, as microglia/macrophages have also receptors for ECM proteins, the dysregulated interstitial ECM in MS lesions also affects microglia/macrophage activation (Austin et al., 2012;Ebert et al., 2008;Milner et al., 2007;Rolls et al., 2008;Sikkema et al., 2018), which may not only affect MMP expression, but also indirectly contribute to OPC differentiation and therefore remyelination (Lloyd & Miron, 2019;Miron et al., 2013). Hence, means to overcome the dysregulated activation of microglia/macrophages in white matter MS lesions (Peferoen et al., 2015;Vogel et al., 2013) may indirectly also reinstall correct ECM remodeling.