Fibrosis in the central nervous system: from the meninges to the vasculature

Abstract

Formation of a collagenous connective tissue scar that forms after penetrating injuries to the brain or spinal cord has been described and investigated for well over one hundred years. However, it was studied almost exclusively in the context of penetrating injuries that resulted in infiltration of meningeal fibroblasts, which raised doubts about translational applicability to most CNS injuries where the meninges remain intact. Recent studies demonstrating the perivascular niche as a source of fibroblasts have debunked the traditional view that a fibrotic scar only forms after penetrating lesions that tear the meninges. These studies have led to a renewed interest in CNS fibrosis not only in the context of axon regeneration after spinal cord injury, but also across a spectrum of CNS disorders. Arising with this renewed interest is some discrepancy about which perivascular cell gives rise to the fibrotic scar, but additional studies are beginning to provide some clarity. Although mechanistic studies on CNS fibrosis are still lacking, the similarities to fibrosis of other organs should provide important insight into how CNS fibrosis can be therapeutically targeted to promote functional recovery.

Introduction

Upon injury, virtually all tissue types mount a wound healing response to initiate tissue repair. This wound healing process typically starts with an inflammatory response by the tissue resident cells that triggers a coagulation cascade to establish hemostasis and infiltration of leukocytes. Cytokines and growth factors expressed by multiple cell types result in cell proliferation to replace dead parenchymal cells. A provisional matrix established mainly by fibrin, fibronectin and collagen provide a source of growth factors and substrate for cell migration. As the matrix is constantly being formed and degraded, activated fibroblasts secrete and interact with the matrix to promote wound contraction. Eventually, the matrix is degraded, and fibroblasts are replaced with parenchymal cells that reestablish functional tissue.

Fibrosis is a pathological wound healing process characterized by functional parenchymal tissue being replaced by non-functional connective tissue with excessive deposition of extracellular matrix (ECM) protein typically secreted by activated fibroblasts as well as other cell types. Excessive fibrosis is a common cause of organ failure, and a leading cause of death in developed countries. Despite being recognized for over 100 years, fibrosis in the central nervous system (CNS) remains understudied and poorly understood compared to gliosis or fibrosis of other organs. This is mostly due to the long-held belief that fibrosis in the CNS was due to infiltration of meningeal fibroblasts after penetrating injuries, which were not common clinical cases. However, recent studies have clearly demonstrated that a fibrotic scar can form in CNS injuries where the meninges remain intact due to the presence of fibroblasts in the perivascular niche. These findings have led to a newfound interest in CNS fibrosis and have opened new possibilities about the formation and contribution of the fibrotic scar in a wide range of CNS injuries.

Historical perspective

The inability of axons to cross through the connective scar tissue that forms after a transection spinal cord injury had been noted since the late 1800’s (Schiefferdecker, 1876, Stroebe, 1894). But one of the first descriptions of the fibrotic scar in the mammalian CNS was by Wilfred Penfield in the 1920’s when he described “formation of a connective tissue core which replaces the injured tissue” surrounded by areas of gliosis after experimental puncture of the brain (Penfield, 1924, Penfield and Buckley, 1928). Penfield noted that this meningocerebral cicatrix formed not only in puncture wounds made by needles but were also present in human epileptogenic brain lesions caused by lacerations, abscess, or meningitis (Penfield and Humphreys, 1940). Ramón y Cajal made similar observations of mesodermal connective scar tissue across a spectrum of CNS lesions in multiple species (Ramón y Cajal, 1928). By observing the presence of sensory axons in the connective scar tissue, he concluded that spinal cord axons “can be powerfully stimulated by means of active or trophic substances liberated by the mesodermic scar and diffused in the spinal wounds and their edges.” On the contrary, “in the cerebellum and cerebrum this vigorous, though ineffective, attempt to innervate the cicatricial connective tissue is always lacking,” noting the limited regenerative capacity of axons in the brain versus sensory axons in the spinal cord. Ramón y Cajal and others have noted that neonatal animals can display significant axon regeneration across the injury site (Ramón y Cajal, 1928, Sechzer, 1974), and later studies showed that this enhanced regeneration in young animals is associated with virtual absence of fibrotic scarring after both brain and spinal cord injuries (Berry, et al., 1983, Kawano, et al., 2005, Li, et al., 2020).

During the next several decades, it became widely recognized that the connective scar tissue forms a formidable barrier against axon regeneration in the spinal cord, and efforts were underway to reduce the fibrotic scar. Windle and colleagues led one of the first efforts by administering Piromen, a pyrogenic bacterial polysaccharide, after spinal cord lesions in cats (Arteta, 1956, Clemente and Windle, 1954). Freeman and colleagues reduced connective scar tissue with local administration of trypsin (Freeman, et al., 1960), as well as transplantation of cultured cerebellar cortical tissue (Kao, et al., 1970) after canine spinal cord lesion. Immunosuppressive treatment using Cytoxan (cyclophosphamide) was used to reduce fibrotic scarring (Feringa, et al., 1974). Matthews et al (1979) repeated the Piromen, Cytoxan, and trypsin studies to provide the first detailed electron microscopy analysis of the fibrotic scar after spinal cord lesions.

The use of more modern methods to reduce fibrotic scarring have also provided mechanistic insight. Berry and Logan and colleagues showed that neutralizing antibodies against TGFβ1 significantly attenuates fibrotic scar tissue after brain injury (Logan, et al., 1994). In a series of studies in both the brain and spinal cord, Müller and colleagues showed that inhibition of collagen synthesis using the iron chelator 2,2’-bipyridine (inhibitor of prolyl-4-hydroxylase) combined with cAMP significantly reduces formation of a fibrotic scar and promotes axon regeneration (Brazda and Müller, 2009, Schiwy, et al., 2009, Stichel, et al., 1999). Raisman and Kawano and colleagues also used an iron chelator (2,2’-dipyridyl) as well as an inhibitor of TGFRβ1 to show reduced fibrotic scarring associated with increased regeneration of nigrostriatal dopaminergic neurons (Kawano, et al., 2012, Kawano, et al., 2005, Yoshioka, et al., 2011). Their findings with a TGFRβ1 inhibitor was consistent with the use of a TGFβ1 antibody mentioned above, and underscored the importance of TGFβ1 signaling in CNS fibrosis.

Until recently, a common feature of all the studies mentioned above as well as virtually all other studies on fibrotic scarring after brain or spinal cord injury was the use of penetrating lesion models that tear the meninges. The source of the fibrotic scar in these penetrating models was shown to be mainly meningeal fibroblasts that infiltrate the injury site; if the injury site is covered with synthetic material that limits meningeal invasion, fibrotic scarring is significantly reduced (Campbell and Windle, 1960). However, applying the material extradurally can also reduce fibrotic scarring, suggesting that fibroblasts from surrounding tissue can also be a major source (Krikorian, et al., 1981). Thus, a limitation of these penetrating injury models was that most clinical CNS injuries maintain meningeal integrity. However, it should be noted that studies have demonstrated fibrotic scarring in severe compressive spinal cord injuries as well as multiple sclerosis lesions in post-mortem human tissue (Mohan, et al., 2010, Norenberg, et al., 2004). Nevertheless, it was commonly believed that only penetrating injuries lead to fibrotic scarring, and that it is largely absent in injuries where the meninges remain intact (Fernandez and Pallini, 1985, Silver and Miller, 2004).

However, this view has been largely debunked by recent discoveries that fibroblasts can originate from the perivascular niche. Frisen and colleagues reported that Type-A pericytes can differentiate into fibroblasts that comprise the fibrotic scar after penetrating spinal cord injury (Goritz, et al., 2011). Lee and colleagues used experimental models of contusive spinal cord injury (Soderblom, et al., 2013) and multiple sclerosis (Yahn, et al., 2020) to show the perivascular origin of fibrosis in non-penetrating CNS injuries. The discovery that the fibrotic scar can originate from the perivascular niche even after CNS injuries where the meninges remain intact have brought renewed interest in CNS fibrosis and potential to identify novel cellular and molecular therapeutic targets.

Perivascular fibroblasts versus pericytes

The first evidence of the perivascular origin of fibroblasts after CNS injury was provided by Goritz et al (2011) in a penetrating model of spinal cord injury. By performing genetic lineage tracing using GLAST-CreER transgenic mice, they demonstrated the presence of a new subtype of pericytes (termed Type A pericytes) that were located abluminal to the more classical pericytes (termed Type B pericytes). Although dorsal column lesion in these mice resulted in a dense fibrotic scar that were comprised of the genetically labeled Type A pericytes, it is not clear whether GLAST-CreER mice also label meningeal fibroblasts that may have contributed to the fibrotic scar after the penetrating injury. To investigate the cellular origin of fibrosis in closed CNS injuries where the meninges remain intact, Soderblom et al (2013) used Col1α1-GFP mice to identify Col1α1-positive perivascular fibroblasts that were located mostly around large diameter vessels in the spinal cord. Performing contusive model of spinal cord injury in these transgenic mice resulted in GFP-positive fibroblasts that filled the fibrotic scar. To address the possibility that pericytes could have upregulated GFP after spinal cord injury, the authors bred Col1α1-GFP mice with NG2-CreER/Rosa26-tdTomato mice, and showed that NG2-positive classical pericytes (consistent with Type B pericytes mentioned above) did not express GFP and were not present in the fibrotic scar after contusive spinal cord injury. Ensuing studies in an EAE (experimental autoimmune encephalomyelitis) model of multiple sclerosis using the same Col1α1-GFP mice showed that a fibrotic scar comprised of perivascular fibroblasts also forms in this model (Dorrier, et al., 2021, Yahn, et al., 2020), further supporting the notion that fibrosis can develop in non-penetrating CNS injuries. Importantly, a recent study by Dorrier et al (2021) used Col1α2-CreER mice and performed genetic lineage tracing studies to unequivocally show that pre-existing Col1-expressing perivascular fibroblasts are the overwhelming majority of the fibroblasts that comprise EAE lesions.

These independent studies have clearly demonstrated the perivascular origin of fibrosis in certain types of CNS injuries, but the cellular nomenclature has led to some confusion and misunderstanding. Is the fibrotic scar comprised of Type A pericytes that differentiate into fibroblasts, or is it comprised of fibroblasts that normally reside in the perivascular space, but migrate to the injury site upon activation by injury, or is it comprised of both populations? The third option is highly unlikely because the authors from the respective studies have shown that their individual cell types comprise virtually all the PDGFRβ-positive cells at the injury site, thereby leaving very little room for multiple cellular origin. One possibility is that these are the same cells, which can be directly addressed in the future by breeding Col1α1-GFP mice with GLAST-CreER/Rosa26-RFP mice and assessing potential overlap between GFP-positive and RFP-positive cells before and after spinal cord injury (similar to the way NG2-CreER mice were used above). Another method is to compare the molecular markers between the two cell types. Type A pericyte markers have been defined as being positive for PDGFRβ, CD13, and negative for desmin (Goritz, et al., 2011). Perivascular fibroblasts have been defined as being positive for Col1α1, PDGFRβ, and CD13, but negative for NG2, nestin, and αSMA (Dorrier, et al., 2021, Soderblom, et al., 2013). However, after injury, activated fibroblasts that comprise the injury site mostly lose their CD13 expression (Soderblom, et al., 2013). Thus, the overlap of PDGFRβ and CD13 expression between Type A pericytes and perivascular fibroblasts suggests that these may be the same cell types, but these are also markers of classical (Type B) pericytes, so more molecular markers need to be assessed before a firmer conclusion can be made.

If these are indeed the same cell type, then does it matter how they are named? If so, which is the more appropriate label? Recent advances in single cell RNAseq studies on the brain vasculature have provided important insight into these questions. Betsholtz and colleagues used single cell RNAseq to perform perhaps the most comprehensive analysis to date of the cells that comprise the brain vasculature (Vanlandewijck, et al., 2018). Using multiple transgenic reporter mice, the authors were able to capture virtually all vascular cells including pericytes and fibroblasts, which formed distinct populations after clustering analysis. Pericytes were defined as being positive for Pdgfrb, Cspg4, Des, and negative for Acta2, Tagln. Fibroblasts were defined as being positive for Pdgfrb, Pdgfra, Col1α1 (and many other collagens), Lum, and Dcn. These unique markers suggest unique signaling pathways in each cell type. In addition to their distinct molecular identifies, pericytes and fibroblasts had different anatomical locations along the vasculature. Whereas pericytes were located mostly around capillaries, fibroblasts were in non-capillary regions that were positive for LAMA1. Clustering analysis showed only one type of pericyte in both the brain and lung, whereas fibroblasts had multiple subtypes. An independent study by Linnarson and colleagues reported similar molecular markers of pericytes and fibroblasts (referred to as vascular leptomeningeal cells) but found three pericyte subtypes (Zeisel, et al., 2018). Although the authors did not go into the details of the three subtypes or compare them to existing pericyte classifications (such as Type A/B), search for GLAST (Slc1a3) in their online database (http://mousebrain.org) showed expression in vascular leptomeningeal cells (i.e. fibroblasts) but not in pericytes. The online database for the Betsholtz study (http://betsholtzlab.org/VascularSingleCells/database.html) shows higher Slc1a3 expression in fibroblasts than pericytes (highest expression is in astrocytes, as expected). Therefore, single cell RNAseq analysis of the brain vasculature clearly shows pericytes and perivascular fibroblasts as two distinct cell types, which means that these two cell labels should not be used interchangeably. More detailed characterization of pericyte heterogeneity is needed to provide a better understanding of the relationship between GLAST-positive Type A pericytes and perivascular fibroblasts. Molecular markers of pericytes and perivascular fibroblasts that have been described in the CNS are listed in Table 1. One difficulty that will remain in validating gene expression data with immunohistochemistry is that many of the unique molecular marker genes, especially for fibroblasts, encode proteins that label ECM rather than the cell itself (such as collagens, lumican, and decorin).

Table 1.

Molecular markers of pericytes and perivascular fibroblasts in the CNS. Common names are in parentheses. Validated markers are in bold.

Pericyte	Pericyte & Fibroblasts	Fibroblasts
Cspg4 (NG2) (Dorrier, et al., 2021, Soderblom, et al., 2013)	Pdgfrb (Goritz, et al., 2011, Soderblom, et al., 2013)	Col1a1 (Soderblom, et al., 2013, Vanlandewijck, et al., 2018)
Des (Desmin) (Dorrier, et al., 2021)	Anpep (CD13) (Goritz, et al., 2011, Soderblom, et al., 2013)*	Col1a2 (Dorrier, et al., 2021)
Kcnj8 (Vanlandewijck, et al., 2018, Zeisel, et al., 2018)		Pdgfra (Vanlandewijck, et al., 2018)
		Dcn (Vanlandewijck, et al., 2018, Zeisel, et al., 2018)
		Lum (Vanlandewijck, et al., 2018, Zeisel, et al., 2018)

^*Soderblom et al showed that CD13 expression is decreased in fibroblasts after injury.

Development of fibrosis across a spectrum of CNS injury

The discovery that CNS fibrosis can arise from the perivascular niche in addition to meningeal fibroblasts, has led to the identification of fibrosis as a pathology common to many different CNS disorders with spinal cord injury serving as a model system. Genetic lineage tracing studies using GLAST-CreER and Col1α1-GFP mice have shown that after spinal cord injury, fibroblasts (or Type-A pericytes) start to accumulate at the lesion site around 3 days post-injury and reach their maximum number at 7 days post-injury (Goritz, et al., 2011, Soderblom, et al., 2013). The dense region of fibroblasts surrounded by the glial scar remains chronically after injury. Genetic lineage tracing studies have demonstrated that the majority of non-vascular PDGFRβ-positive cells in the lesion are fibroblasts, which has led to a more accessible method of using an antibody against PDGFRβ to label fibroblasts in the scar, especially in non-transgenic animals. This histological approach has shown that a fibrotic scar forms and persists in rat models of contusive spinal cord injury (Zhu, et al., 2015b), which historically were thought to not develop a major fibrotic scar. Notably, a proteomics study of the rat contusion injury site showed Col1α1 as one of the most highly expressed ECM along with several other fibrotic markers (Didangelos, et al., 2016). Despite the fact that a large portion of the fibrotic scar turns into a cystic cavity in rats (species differences discussed more below), these studies are important because compared to mice, rat spinal cord injury pathobiology is considered to be more similar to humans. Therefore, the use of better molecular markers and methods to identify fibrosis in rat studies raise the need to perform careful analyses of fibrotic scar formation in human post-mortem spinal cord injury tissue.

Recent studies show that fibrosis is not unique to traumatic injuries such as spinal cord injury. Post-mortem analysis of human multiple sclerosis tissue revealed the deposition of perivascular fibrotic ECM (e.g., Col1α1), emphasizing the clinical relevance of fibrosis in multiple sclerosis (Brown and Thore, 2011, Mohan, et al., 2010). Using EAE model of multiple sclerosis, Yahn et al (2020) used Col1α1-GFP mice to show that a fibrotic scar derived from perivascular fibroblasts is closely associated with demyelinated regions of the spinal cord. Interestingly, unlike spinal cord injury, EAE fibrotic scars partially resolved over time (size of fibrotic scar became smaller at chronic stages). Furthermore, the authors used in vitro co-culture assays to show that both fibroblast conditioned media and ECM can inhibit oligodendrocyte differentiation. A more recent study confirmed the fibrosis pathology after EAE in Col1α1-GFP mice, but also used Col1α2-CreER mice to perform genetic lineage tracing studies using the gold standard inducible cre-mediated recombination (Dorrier, et al., 2021). By breeding the two mouse lines together, the authors found that the genetically labeled cells in Col1α2-CreER mice were the same as Col1α1-GFP perivascular fibroblasts, and that the genetically labeled Col1α2 cells comprised the vast majority of fibroblasts in the lesion. Although this study needs to be performed in other injury models such as spinal cord injury, it is the strongest proof yet that fibroblasts in the fibrotic scar after CNS injury arise predominantly from pre-existing perivascular fibroblasts rather than other cell types that may have upregulated collagen 1 expression after injury. Furthermore, genetic ablation of proliferating fibroblasts reduced severity of EAE, although it did not improve area of myelination.

Fibrosis has been observed in many other non-penetrating CNS injuries. In an MCAO (middle cerebral artery occlusion) model of stroke in mice and rats, PDGFRβ-positive cells as well as areas of collagen deposition were found to be present at the lesion site (Fernandez-Klett, et al., 2013, Iihara, et al., 1996, Makihara, et al., 2015). In post-mortem brain tissue from stroke patients, PDGFRβ-positive cells were also found within the lesion site surrounded by GFAP-positive astrocytes (Fernandez-Klett, et al., 2013). In addition, following an optic nerve crush in Col1α1-GFP mice, perivascular fibroblasts form a fibrotic scar at the injured optic nerve (Liu, et al., 2021).

In addition to the cellular components of the fibrotic scar, the ECM plays a significant role in the lesion pathobiology by serving not only as a physical substrate for cell adhesion, but also binding and regulating the properties of cytokines, growth factors and chemorepulsive guidance molecules (de Winter, et al., 2016, Schultz and Wysocki, 2009, Xu and Clark, 1996). One of the best characterized ECM components common to all the CNS fibrotic lesions mentioned above is fibronectin. In addition to fibroblasts, other sources of fibronectin include macrophages, astrocytes, and endothelial cells (Zhou, et al., 2019). To function properly, fibronectin and other ECM such as collagen must assemble into a matrix. Acutely after spinal cord injury, fibronectin is present in a soluble form (most likely as plasma fibronectin), and by 7 days post-injury assembles into a matrix of cellular fibronectin. Although this matrix persists chronically, the amount of matrix fibronectin significantly decreases over time (Zhu, et al., 2015b). Genetic ablation of an isoform of fibronectin containing the Extra Domain A domain (FnEDA) led to significantly reduced fibronectin deposition chronically after spinal cord injury, which was associated with reduced lesion size, higher axonal density in the lesion, and improved locomotor function (Cooper, et al., 2018). In post-mortem multiple sclerosis tissue, fibronectin aggregates were present in chronically demyelinated lesions (Stoffels, et al., 2013). Genetic deletion studies in mice showed astrocytes to be the major source of fibronectin in demyelinated lesions, and injection of astrocyte-derived fibronectin in toxin-induced demyelinated lesions inhibited oligodendrocyte differentiation and remyelination (Stoffels, et al., 2013, Stoffels, et al., 2015). Taken together, these studies indicate that fibronectin contributes to the CNS injury site environment that limit a regenerative response.

Mechanisms of CNS fibrosis

One of the first mechanistic insight into CNS fibrosis was provided by the discovery that intracerebroventricular injection of a neutralizing antibody against TGFβ1 reduced, whereas injection of recombinant TGFβ1 enhanced, fibrotic scarring after a penetrating brain lesion in rats (Logan, et al., 1994). Furthermore, pharmacological inhibition of TGFβR1 also reduced fibrotic scar formation after a penetrating brain lesion (Yoshioka, et al., 2011). The source of TGFβ1 has been reported to be cells of the microvasculature, astrocytes, microglia, and macrophages in both rat brain injury and human post-mortem spinal cord injury tissues (Buss, et al., 2008, Logan, et al., 1994), and both TGFβR1 and TGFβR2 are expressed by meningeal fibroblasts after penetrating brain lesions (Komuta, et al., 2010). Thus, TGFβ1-TGFβR signaling axis seems to contribute significantly to fibrosis after brain injury. It remains to be seen whether this is a pathway common to other CNS injuries that result in a fibrotic scar, but it seems highly likely since TGFβ1 is a canonical signaling pathway in many other organs (Meng, et al., 2016). Based on these pharmacological studies, we still do not understand the network of cell-to-cell interactions responsible for TGFβ1-mediated fibroblast activation after CNS injury, but future cell-type specific genetic knockout studies can provide further mechanistic insight.

One cell type that has been shown to contribute to CNS fibrosis is monocyte-derived macrophages. Acute macrophage depletion using clodronate liposomes led to a significant reduction in the number of fibroblasts that were present at the injury site, which was associated with smaller lesion size and increased number of neurofilament-positive axons in the lesion (Zhu, et al., 2015a). Terminating clodronate liposome administration led to reinfiltration of macrophages and reestablishment of the fibrotic scar, suggesting that macrophages contribute to fibrosis at both the acute and chronic phases of spinal cord injury (Zhu, et al., 2015a). Furthermore, intravenously injected poly(lactide-co-glycolide) nanoparticles were phagocytosed by blood monocytes, which inhibited their migration to the spinal cord injury site resulting in decreased fibrosis, improved histopathology, and locomotor recovery (Jeong, et al., 2017). In addition, regulated release of an anti-inflammatory drug (methylprednisolone) via a novel porous nanoscaffold resulted in decreased macrophage infiltration and pro-inflammatory cytokine expression, which were associated with reduced fibrosis after spinal cord injury in mice (Yang, et al., 2020). The molecular signaling between macrophages and fibroblasts during CNS fibrosis is currently unknown, but it is likely to be similar to mechanisms described in other organ systems (Distler, et al., 2019). For example, as described above, TGFβ1 is an important regulator of fibrosis, and TGFβ1 and its receptors are expressed by macrophages and fibroblasts respectively in the injured CNS (Buss, et al., 2008, Komuta, et al., 2010, Logan, et al., 1994). However, macrophages are known to express other profibrotic factors such as PDGF (platelet derived growth factor), and accordingly, fibroblasts express both PDGFRα and PDGFRβ (Dorrier, et al., 2021, Vanlandewijck, et al., 2018). Thus, attenuation of fibrosis after macrophage depletion is likely due to reduction in several profibrotic molecules normally expressed by macrophages after injury.

In addition to the innate immune response, adaptive immunity also seems to contribute to CNS fibrosis. A recent study demonstrated that genetic deletion of IFNγR1 (Interferon gamma receptor 1) specifically in fibroblasts reduces fibrosis after EAE (Dorrier, et al., 2021). Since overexpression of IFNγ itself did not promote fibrosis, the data suggest that IFNγ is contributory but individually insufficient to induce fibrosis after EAE. The source of IFNγ based on single cell RNAseq analysis was determined to be T cells, which highlights the contribution of the adaptive immune response in fibrosis after EAE. The above studies highlight the importance of the inflammatory response from both the innate and adaptive immune systems in development of fibrosis after CNS injury, however, the degree of neuroinflammation required to induce fibrosis is not clear. Spinal cord injury and multiple sclerosis result in high level of neuroinflammation with a severe breakdown of blood-CNS barrier. Whether fibrosis will develop in injury models such as systemic LPS injections or in neurodegenerative disorders with a relatively more subtle inflammatory response remains to be seen.

The fibrotic scar seems to interact with neural cells to create an environment inhibitory to regeneration. Neurons grown on astrocyte/fibroblast co-culture cannot extend axons from the astrocyte to the fibroblast side very well (Shearer and Fawcett, 2001). Oligodendrocyte progenitor cells treated with fibroblast conditioned media or grown on fibroblast ECM display significantly reduced capacity to differentiate into mature oligodendrocytes in vitro (Yahn, et al., 2020). Type I collagen produced by fibroblasts induces astrogliosis through the N-cadherin pathway, and inhibition of astrocyte-type I collagen interaction reduces astrogliosis and enhances axon regrowth (Hara, et al., 2017).

Although mechanisms of CNS fibrosis are still poorly understood, it is becoming evident that they share many similarities with how fibrosis develops in other organs. Like the CNS, inflammation is the major driver of fibrosis in most organs (Mack, 2018, Wynn, 2008). Furthermore, many of the same profibrotic signaling pathways in the CNS have been identified in other organs. In addition to TGFβ1 as discussed above, Wnt, hedgehog, and interleukin family members are also canonical profibrotic pathways that could also contribute to CNS fibrosis (Wynn, 2008). In fact, spinal cord injury in Wnt reporter mice showed Wnt signaling activation in cells that comprised the fibrotic scar (Yamagami, et al., 2018). Importantly, the overlapping signaling pathways implies that the intensive effort by the biotech industry to develop anti-fibrotic agents in fibrosis of other organs could lead to discovery of therapeutics for CNS indications where fibrosis is a major pathology.

Targeting fibrosis for CNS repair

As discussed above, several approaches targeting CNS fibrosis have already been reported in numerous studies. These include neutralizing antibody against TGFβ1, small molecule inhibitor of TGFβR1, and iron chelators inhibiting prolyl hydroxylase. The initial studies used these approaches more as tools than as therapeutics, but it is interesting to note the availability of therapeutics with similar targets that are currently available or being tested in the clinic. FDA-approved iron chelators, such as deferoxamine (Desferal), are typically used to treat iron overload syndromes, but there is accumulating evidence that it reduces fibrosis in multiple disease indications (Ikeda, et al., 2014, Mohammed, et al., 2016, Shen, et al., 2020). Accordingly, when given after penetrating spinal cord injury in rats, deferoxamine has been shown to reduce fibrosis and slightly improve behavioral recovery (Vogelaar, et al., 2015). Furthermore, deferoxamine has been used in multiple independent studies as a neuroprotective agent since iron overload affects many pathological processes (Hao, et al., 2017, Liu, et al., 2011). Another notable antifibrotic is pirfenidone (Esbriet), which is currently approved to treat idiopathic pulmonary fibrosis. Pirfenidone’s molecular target(s) are not clear, but it has several mechanisms of action including reducing expression of TGFβ1 as well as TNF and Il-1β (Ruwanpura, et al., 2020). Although pirfenidone has not yet been tested as an antifibrotic after CNS injury, it has shown some positive results as a neuroprotective agent in progressive multiple sclerosis clinical trials (Walker, et al., 2005). It would be interesting to investigate whether pirfenidone and other FDA-approved TGFβ1 inhibitors have antifibrotic effects in animal models of CNS injury.

One of the best known antifibrotic in models of spinal cord injury is the microtubule stabilizing chemotherapeutic paclitaxel (Taxol). Administration of Taxol to the injury site through an intrathecal catheter resulted in significant reduction in fibrosis after models of penetrating as well as contusive spinal cord injury (Hellal, et al., 2011, Popovich, et al., 2014). The proposed mechanism is that stabilizing the microtubule network interferes with TGFβ1 signaling by preventing Smad2/3 nuclear translocation. Since Taxol had to be delivered locally due to its inability to penetrate the blood-spinal cord barrier, later studies used the blood-spinal cord barrier permeable FDA-approved microtubule stabilizing agent epothilone B. Systemic (intraperitoneal) injections of epothilone B significantly reduced fibrotic scar formation, enhanced axon regeneration and locomotor recovery after contusive spinal cord injury in rats (Ruschel, et al., 2015). Systemic injections with epothilone D, which has a higher therapeutic index than epothilone B, also has similar effects, albeit more muted than compared to epithilone B (Ruschel and Bradke, 2018, Sandner, et al., 2018).

For intuitive reasons, the general strategy for targeting fibrosis, whether in the CNS or in peripheral organs, has been to reduce it. However, genetic ablation studies have brought new insight to this approach. Ablation of dividing fibroblasts after EAE reduced the fibrotic scar and improved motor recovery (Dorrier, et al., 2021), which is consistent with the expectation that reducing fibrosis is beneficial. However, fibroblast ablation studies after spinal cord injury have also led to a more complex understanding of the fibrotic scar. After a penetrating spinal cord injury, inhibiting proliferation of Type A pericytes led to either a reduction in the fibrotic scar if the recombination efficiency of the transgenic mouse was moderate or the presence of a large fluid-filled cystic cavity in lieu of the fibrotic scar if the recombination efficiency was high (Dias, et al., 2018). In an independent study using a different mouse line, almost completely preventing fibrotic scar formation after contusive spinal cord injury led to loss of tissue integrity of the injury site, and worse behavioral outcome (Hesp, et al., 2018). These studies indicate that after spinal cord injury, completely preventing fibrotic scar formation leads to improper wound healing at the injury site that is detrimental to tissue repair and behavioral recovery. One intriguing explanation for these beneficial and detrimental roles of the fibrotic scar is the possibility of different types of fibroblasts; one that is associated with repair versus one that is inhibitory to regeneration. This is a possibility worth pursuing in future studies.

The above ablation studies were all performed in mice due to their genetic tractability, but it is important to recognize species differences in spinal cord injury pathology between mice and rats. After a penetrating spinal cord injury, both species develop a dense collagenous fibrotic scar comprised mainly of meningeal fibroblasts that invade the lesion. However, after a more clinically relevant model of contusive injury where the meninges remain largely intact, a large portion of this fibrotic scar in rats and humans (but not in mice) turns into a cystic cavity very similar to what has been observed after fibroblast ablation in mice. Although the reasons for this difference are unclear, it is most likely not due to differences in spinal cord size between mice and rats (Inman and Steward, 2003). Thus, in the context of wound healing, we can speculate that, after a contusive spinal cord injury, mice represent an over healing response that develops into a fibrotic scar whereas rats represent a failure of the wound healing response that results in an open wound (i.e., cavity). A logical extension of this speculation is that for CNS injuries that result in a dense fibrotic scar (e.g., penetrating injuries or multiple sclerosis), therapeutic strategies that reduce fibrosis are warranted, but injuries that result in cavitation (e.g. contusive spinal cord injury) may warrant strategies that improve wound healing rather than reducing fibrosis. This strategy is supported by a previous study demonstrating that injection of a poly(organophosphazenes) hydrogel into the spinal cord injury cavity results in a fibrotic scar-like environment that virtually abolishes cavitation chronically and improves white matter sparing as well as locomotor function (Hong, et al., 2017). Interestingly, reducing macrophage infiltration into the hydrogel or adding Taxol abolished the positive effects of the hydrogel and led to cavitation.

In summary, CNS fibrosis is no longer a pathology limited to penetrating injuries that promote infiltration of meningeal fibroblasts. In closed injuries, such as EAE or stroke, perivascular fibroblasts can form the fibrotic scar, and this process is mediated by an inflammatory response from infiltrating peripheral leukocytes. Molecular mechanisms underlying CNS fibrosis seem to share many similarities with fibrosis of other organs, which suggest that the intense efforts in therapeutic development for fibrotic disorders such as pulmonary fibrosis may also aid in finding a treatment for CNS fibrosis. However, genetic ablation studies indicate that CNS fibrosis, similar to astrogliosis, has some neuroprotective effects, and that we need to understand it in a larger context of wound healing in order to develop an appropriate treatment strategy.

Figure 1. Schematic of fibrosis after CNS injury.

In physiological conditions (left vasculature), perivascular fibroblasts line large diameter vessels whereas pericytes are mostly around capillaries. In areas of overlap between the two cell types (circled region), perivascular fibroblasts are located abluminal to pericytes that directly surround endothelial cells. After injury (right vasculature), perivascular fibroblasts migrate away from the vasculature, and proliferate at the injury site. This process is mediated by infiltrating leukocytes (such as macrophages and T cells), which express cytokines such as interferon gamma (IFNγ) and transforming growth factor beta 1 (TGFβ1) that activate fibroblasts to express extracellular matrix (ECM) molecules that are deposited at the injury site.