The Biology of Regeneration Failure and Success After Spinal Cord Injury

Abstract

Since no approved therapies to restore mobility and sensation following spinal cord injury (SCI) currently exist, a better understanding of the cellular and molecular mechanisms following SCI that compromise regeneration or neuroplasticity is needed to develop new strategies to promote axonal regrowth and restore function. Physical trauma to the spinal cord results in vascular disruption that, in turn, causes blood-spinal cord barrier rupture leading to hemorrhage and ischemia, followed by rampant local cell death. As subsequent edema and inflammation occur, neuronal and glial necrosis and apoptosis spread well beyond the initial site of impact, ultimately resolving into a cavity surrounded by glial/fibrotic scarring. The glial scar, which stabilizes the spread of secondary injury, also acts as a chronic, physical, and chemo-entrapping barrier that prevents axonal regeneration. Understanding the formative events in glial scarring helps guide strategies towards the development of potential therapies to enhance axon regeneration and functional recovery at both acute and chronic stages following SCI. This review will also discuss the perineuronal net and how chondroitin sulfate proteoglycans (CSPGs) deposited in both the glial scar and net impede axonal outgrowth at the level of the growth cone. We will end the review with a summary of current CSPG-targeting strategies that help to foster axonal regeneration, neuroplasticity/sprouting, and functional recovery following SCI.

I. INTRODUCTION

When asked in an anonymous questionnaire whether emergency care professionals would desire resuscitative measures following a severe spinal cord injury (SCI), only 37% of physicians and 14% of nurses indicated that they themselves would want measures taken to ensure survival (138). This bleak perception among health care providers of life following SCI derives not only from the devastating permanence of such an injury, but also underscores the dearth of available treatments. In the time that has elapsed since this survey was conducted in the 1990s, our understanding of the molecular mechanisms underlying regenerative failure after SCI has greatly improved, providing hope for novel treatments. Indeed, SCI research has come a long way from Ramon y Cajal’s descriptions of largely abortive regeneration of the “sterile clubs” of severed axonal tips transiently stalled in the glial scar which he believed were destined to die away (56). David and Aguayo’s (81) peripheral nerve graft experiments provided an important appreciation that certain types of central nervous system (CNS) axons do possess some level of an intrinsic ability to regenerate, at least acutely after injury, when presented with a hospitable environment. This galvanized new hope that additional SCI research could lead to functional improvements.

There are ~17,000 new cases of SCI recorded each year in the United States with motor vehicle collisions still contributing to the most common etiology of traumatic cord injury (279a). Alarmingly, since 2012 as the population ages, the number of recorded cases of traumatic SCI due to falls has risen from 16 to 30.5% (92). Clinical presentation is dependent on the severity and location of injury since the long tracts within the cord are arranged somatotopically as they interconnect the brain with autonomic and peripheral nervous systems. Most current acute treatments aim to stabilize the cord and restore homeostasis immediately following injury while long-term treatments mainly seek to manage symptoms arising from maladaptive plasticity and other secondary complications (65, 101, 199).

Our understanding of the extrinsic and intrinsic factors that block axonal regeneration or neuroplasticity has vastly improved with greater elucidation of the course of molecular events that occur following SCI in animal models. However, we need to keep in mind that variation of inflammatory and subsequent glial responses exists between mice and rats (365), different genetic mouse backgrounds (202, 243), and rat strains (312). There is variability in the mechanics of the different injury models as well as deviations in severities of the same injury type (369). In this review, we describe how primary and subsequent secondary injuries contribute to the formation of the glial scar (FIGURE 1). Major cellular components of the glial scar including microglia and peripherally derived leukocytes, astrocytes, oligodendrocyte progenitors, ependymal cells, and pericytes/fibroblasts will be further described in detail. We then discuss how the matrix contents of the glial scar, notably chondroitin sulfate proteoglycans (CSPGs), inhibit axon outgrowth and functional recovery. We also review the inhibitory role of CSPGs in the perineuronal net (PNN). Given that the molecular and cellular cascade of events and possible repair strategies are so expansive, we focus our review from our unique perspective based on modulation of the glial scar by mitigating the inhibitory potential of CSPGs, as well as combinatorial strategies that target the glial scar.

FIGURE 1.

Overview of spinal cord injury pathophysiology. Spinal cord injury can be divided among four progressive stages: physical trauma, primary injury, secondary injury, which ultimately creates a chronically axon-inhibitory structure called the glial scar.

II. DESCRIPTION OF SCI-INDUCED PRIMARY AND SECONDARY INJURY

In rodent models of contusive SCI, the vertebrae and surrounding laminae are excised to expose the cord. The desired amount of force is translated through a clip, piston, or a set amount of weight that is impelled or dropped onto the protracted cord, which causes immediate damage that expands in an elliptical shape rostrally and caudally from the center of impact. While there is some variability in the extent of spared tissue, this model has been useful in recapitulating major morphological and cellular changes present in traumatic human SCI (369). In the contusive model of injury, the force of impact rapidly displaces the cord (FIGURE 2, A–C). While physical trauma is usually brief, tissue remodeling starts immediately, peaks weeks after injury, and persists for the remainder of life.

FIGURE 2.

Primary injury. Primary injury caused by direct trauma induces blood spinal cord barrier (BSCB) disruption among other disruptions including ischemia and reperfusion injury. A–C: single (A), multi-level edema (B), and hemorrhage and surrounding edema (C) visualized in human spinal cord injury seen through sagittal T2 MRI with C1–C6 injuries. (From Bozzo et al. J Neurotrauma 28: 1401–1411, 2011. Copyright Mary Ann Liebert, Inc.) D: traumatic injury induces hemorrhaging following BSCB disruption as seen in a human case of C4–5 at 3 days following SCI. (From Tator and Koyanagi. J Neurosurgery 86: 483–492, 1997. TheJNS.org) E and F: ischemia and reperfusion injury result in neuronal and glial apoptosis marked by dense cellular condensation (E) and necrosis marked by cytoplasmic blebbing (F) as seen through EM imaging of the spinal cord. [From Liu et al. (237), with permission from Society for Neuroscience conveyed through Copyright Clearance Center, Inc.]

A. Physical Trauma and Primary Injury

Physical impact to the spinal cord causes symmetrically expanding foci of glial and neuronal necrosis within the lesion epicenter (FIGURE 2F) (146). Necrosis is a disordered phenomenon of cell death characterized by somal swelling, loss of cytoplasmic definition, chromatin aggregation that may leak into the cytoplasm, and spewed organelles resulting in ejection of proteins that contribute to the inflammatory milieu (18). These include the release of alarmins that initiate a reactive state in resident glia and the downstream infiltration of immune cells from the periphery (FIGURE 3) (33).

FIGURE 3.

Secondary injury. Secondary injury is marked by inflammation initiated by physical injury and release of alarmins. A–C: leukocyte infiltration through a compromised BSCB and subsequent differentiation can be seen in the lesion of the mouse spinal cord after injury. While both M1 (CD16/32+) and M2 (arginase 1) macrophages are present, the M1 type dominates and persists. [From Kigerl et al. (201), with permission from Society for Neuroscience conveyed through Copyright Clearance Center, Inc.] D and E: M1-like macrophages may induce further damage to axons through dieback, or contact-based retraction of the axon to the soma as seen with peripheral rat axons in vitro. However, immune modulatory treatments such as MAPC stem cells may push macrophages towards an M2-like phenotype which does not cause dieback in vitro even upon axon contact. [From DePaul et al. (89), with permission from Nature Publishing Group.]

In contrast, apoptosis is an ordered, ATP-driven process of cell death marked by tightly packed condensation of the cell body that elicits a far less potent inflammatory signal than necrosis (FIGURE 2E) (209). Neuronal apoptosis seems confined to the early stages of injury (147). However, beyond 24 h post trauma, glial apoptosis persists while neuronal apoptosis diminishes (74, 237). Among glia, astrocytes seem most refractory to apoptosis (146) while oligodendrocyte apoptosis persists in distant white matter several days or longer after impact (29, 237, 374).

Physical impact additionally causes descending and ascending axons along the lesion to shear and then degenerate (277). These severed axons, separated from their somata, undergo stereotypical Wallerian degeneration of the distal aspect of the axon as well as their associated myelin sheaths. Successive axonal dieback (49, 171) due to aggressive macrophage activity (70, 201) and demyelination (41) also occur in the proximal segment which, in turn, may contribute to the list of factors that limit collateral sprouting (75).

Understanding the homeostatic and protective functions of the blood-spinal cord barrier (BSCB) in keeping peripheral immune cells, toxic metabolic products, and other inflammatory substances excluded from the CNS emphasizes the extent of disruption that SCI incurs. The BSCB comprises nonfenestrated endothelial cells which form tight junctions with each other to wrap around blood vessels. The next layer includes endothelial cell-supporting pericytes embedded in the basal lamina which are all enveloped by the endfeet of astrocyte processes (182) that play a major role in tightening the endothelial barrier (24). Upon injury, endothelial cells shed their glycocalyx, and tight junctions are lost leading to breakdown of the BSCB and vascular permeability (255). Further disruption of this coordinated structure leads to edema as a result of the unfettered flow of ions and water. Of course, a major complication of BSCB and blood vessel rupture is hemorrhaging (FIGURE 2D), notably in the grey matter (41, 211). Bleeding occurs within minutes after SCI (18), but can last for days (265) contributing to the cavitation of the lesion site which can further potentiate expansion of the injury area by causing compression further along the cord (393). Neuronal excitotoxicity from the release of ions such as Ca²⁺ and glutamate also occurs (3, 4). Adding to the homeostatic imbalance following edema includes the unchecked release of neurotransmitters, K⁺, and Na⁺ (229) that can continue to drive neuronal oxidative stress, protein aggregation, and lipid peroxidation (132).

An immediate response to staunch the invading flow of blood into the cord is the release of platelet-derived factors that cause vasospasm as well as endothelin-induced constriction of blood vessels, leading to ischemia in the cell-dense area of the grey matter (333, 334). Oxidative stress due to ischemia is another cause of glial apoptosis occurring as soon as 3 h at the lesion epicenter (277). Hypoxia also facilitates expansion of the lesion (211, 382) and contributes to patchy areas of necrosis near the lesion site (18, 41, 414). There is recent evidence that a low-grade ischemia develops slowly but persists chronically in the caudal aspects of the lesioned cord due to the action of trace amines on pericyte contraction (235). Early after injury, the reintroduction of oxygen to ischemic areas instigates reperfusion injury adding to oxidative damage and the release of reactive oxygen species (ROS) that may further contribute to glial apoptosis and inflammation (160). Reperfusion injury also introduces leukocytes to the cord which may inundate the area with other proinflammatory factors such as cytokines and interleukins. As injury resolves, multiple cystic cavities may form in the lesion site with greater incidence if the dura has opened (328). At this point, the lesion epicenter is replete with red blood cells, cellular debris including myelin and subcellular components, necrotic and apoptosing neurons and glia, all of which contribute to the initiation of secondary injury.

Adding to injury is the unregulated influx of peripherally circulating proteins and factors including ROS, inflammatory factors such as tumor necrosis factor (TNF)-α and transforming growth factor (TGF)-β, nitric oxide, and fibrinogen (319) [see Garcia et al. (132) for a full review]. While fibrinogen is integral to the formation of blood clots to help resolve hemorrhaging, it also helps initiate astrogliosis (327) and activates microglia (183). Infiltrating ROS and other factors such as inducible nitric oxide synthase (iNOS) contribute to neuronal and glial apoptosis in the first 24 h (424). Even as homeostasis and subsequent BSCB repair occurs, the basal lamina of the BSCB remains damaged, and in the absence of astrocytes in the lesion core, the junctions between endothelial cells fail to reform to their pre-injury state due to decreased expression of tight junction proteins such as claudins and occludens (73). Thus the profuse vasculature within the lesion remains highly abnormal, although much of it is likely resorbed eventually leading to formation of the lesion cavity. Activated endothelial cells also potentiate secondary injury through the release of potent chemoattractants such as interleukin (IL)-16 (275). As a result, the BSCB remains chronically leaky after SCI (285). Another contributor to leakiness is the continued expression of matrix metalloproteases (MMPs), especially MMP-9 from chronically activated endothelial cells (225), oligodendrocyte progenitor cells (343), reactive astrocytes (438), and possibly pericytes (399). Persistent permeability of the BSCB allows for the chronic infiltration of monocytes into the cord (30, 202), which further potentiates the inhibitory effects of the lesion (49).

B. Inflammation Induces Secondary Injury

Primary injury leads directly to a prolonged secondary injury cascade which lasts for weeks until the wound seals and the glial scar matures. Secondary injury is marked by the expansion of tissue damage from the lesion epicenter including inflammation-induced apoptosis of adjacent cells and axotomy of neurons that may have survived the initial impact (115). The magnitude of inflammation and subsequent secondary injury correlates with the extent of immune cell reactivity observed in response to danger-associated molecular patterns or DAMPs (323). DAMPs in the context of SCI include endogenous alarmins, which are normally cell-bound proteins such as ATP, chromatin-associated protein HMGB1, S100, histones, or interleukins such as IL-1α released from cells undergoing necrosis and apoptosis. DAMPs also include cellular debris such as material shed from sheared axons [see Gadani et al. (126) or Kigerl et al. (200) for a full review of DAMPs following CNS injury].

As the name suggests, alarmins serve as a siren to awaken the immune system, initiating and perpetuating a sterile inflammatory response in the CNS (FIGURE 3) (33). ATP is one example of an alarmin released by injury to the CNS that initiates a rapid polarization of proximate microglia (80), which in turn further potentiates inflammatory signals by producing chemokines such as CCL3 (195) and inducing CCL2 release by reactive astrocytes (297). HMGB1, a protein normally sequestered in the nucleus with chromatin, is another classic alarmin that is released quickly following SCI due to necrosing neurons and has been reported to persist chronically after the resolution of the initial inflammatory response (301). HMGB1 activates and further potentiates inflammation by increasing the release of chemokines such as CXCL1/2 from astrocytes (306). Immune cells of the spinal cord become classically activated to assume an M1 phenotype in this environment, which in its simplest terms is marked by iNOS and increased ROS and reactive nitrogen species (FIGURE 3, A–C). This is a change from homeostatic alternatively activated, or M2 phenotypes which describes a heterogeneous population of immune cells that are marked by Arg1 [see Cherry et al. (70) for a detailed discussion]. For example, microglia become activated by the release of alarmins as well as hemorrhaged material, such as iron disgorged from lysed red blood cells. Phagocytosis of iron by monocytes (332, 338) could contribute to the inflammatory milieu through classical activation of microglia (212). Recently, Gadani et al. (127) reported the release of the alarmin IL-33 by oligodendrocytes and grey matter astrocytes in response to SCI. Genetic knockout of IL-33 in mice resulted in the decreased recruitment of monocytes and a reduction in the inflammatory response (127), further demonstrating that alarmins contribute to recruitment, activation, proliferation, and differentiation of peripheral monocytes in the lesion area (87, 126). An understanding of the cascade of molecular events that occur soon after injury can help lead to improved neuroprotective strategies in an attempt to save as much nervous tissue as possible.

C. Microglia

Under normal physiological conditions, microglia are constantly patrolling the CNS by employing an incessant retraction and extension of processes (14, 80, 283). As such, microglia serve a housekeeping function by removing debris, and also act as a first barrier of defense against invading pathogens and injury (287). Microglia are biased toward an M2-like phenotype during normal brain function (70). Following injury, microglia rapidly change their morphology by initially extending their processes toward the damaged site (80). Soon thereafter, they adopt an amoeboid shape (415) before migrating to the lesion area in response to alarmins such as IL-33, IL-1B, and TNF-α released as soon as 15 min after the lesion (127, 323). The acute inflammatory environment post-SCI, therefore, pushes microglia towards a M1-like bias (151, 201) that contributes to a further loss of neurons and increased astrogliosis in discrete regions (405). Histological studies have found a concentration of microglia very early on in the lesion core (143, 326, 344) followed by their migration to the lesion’s margins later on (105). Activated microglia are characterized by deramified and shortened processes that extend rostral and caudal to the lesion (156). At the lesion, microglia clear debris (143), which is all the more important as apoptosis of oligodendrocytes and neurons continues (63, 347, 364). In response to the inflammatory signals released within the lesion, microglia proliferate profusely (143). Proliferation peaks between 3 and 7 days in the lesion epicenter (365) and plateaus 2–4 wk post injury (312). During this time, although microglia may play a beneficial role by clearing cellular debris from the injury site and helping to seal and block the spread of the lesion (167), they have also been reported to contact damaged axons and phagocytose dendrites (143, 415), which may exacerbate synaptic damage. The resolution of microglial activation as the scar forms requires feedback from astrocytes which includes increases of TGF-β and IL-4 and a simultaneous decrease in proinflammatory cytokines such as IL-1β, TNF, and IL-6 (287). However, well after maturation of the glial scar, low-level microglial activation chronically persists in the brain to affect cognitive function (417) as well as in deafferented areas rostral and caudal to the lesion where it may be involved with upregulation of extracellular matrix (ECM) components and circuit remodeling as well as the production of neuropathic pain (91, 151, 156, 415).

D. Leukocytes

Following BSCB damage and subsequent neutrophil infiltration, monocytes influx into the cord where they differentiate into macrophages in the proinflammatory milieu (281). Through a series of stereotyped processes, collectively called extravasation, leukocytes perform rolling adhesion and then transmigrate through the BSCB where they travel towards the source of the chemoattractant (370, 396). Neutrophils are the first reported peripheral immune cell to infiltrate the spinal cord by 3–6 h post injury (66, 381), although their presence in the cord is short lived (117). The amount of neutrophils present in the cord peaks around 1 day post injury where they are associated with necrotic regions (60). At the lesion site, neutrophils help potentiate the inflammatory cascade (66, 145) by activating other immune cells and glia through the release of proinflammatory cytokines and chemokines (60, 371, 372, 396), ROS, and proteases (159).

Extravasated monocytes differentiate into activated macrophages in the injured cord (201) during two peaks: the first at 1 wk post injury, then at 60 days post injury where they have been reported to persist up to 40 days after differentiation (312). M1-activated macrophages dominate the early lesion site (201) and initiate secondary damage through the secretion of enzymes and proinflammatory factors and by potentiating apoptosis (197). Activated immune cells additionally contribute to neuronal and glial cell death by generating ROS through increased NADPH oxidase activity (132, 288). The complex interactions between different cell types of the lesion epicenter further potentiate inflammation. For example, activated oligodendrocyte progenitor cells may further activate macrophages (325) and vice versa (215, 322, 338, 416). The release of TNF-α from activated microglia also activates astrocytes (219), which discharge glutamate to cause neuronal excitotoxicity and further inflammation (132).

Recruited to the core and margins of the lesion (344), macrophages prolifically phagocytose cellular debris, and degenerating tissue remains (143). Macrophages proliferate in greater numbers than microglia and are characterized by their rounder, more amoeboid shape (105). M1 macrophages have been reported to be the dominant macrophage type found at the lesion (201), although other macrophage subtypes based on their chemokine cell-surface receptors are also present (105). M1-promoting genes have been reported to increase up to a month post injury while M2-promoting genes are depressed in the early injury environment (201). Far more than microglia (105), macrophages make intimate contact with the dystrophic ends of axons which promotes axonal retraction or dieback (FIGURE 3D). The dying back phenomenon appears to be mediated, at least in part, by MMP-9 (49, 105, 171). Dieback is terminated once the axonal retraction bulb makes sufficient contacts with NG2⁺ oligodendrocyte progenitor cells and possibly pericytes where the dystrophic tip makes synaptic-like contacts and stabilizes upon these glial progenitor cells (see below). Thus activated macrophages in the early stages of the development of the lesion core play a major role in regeneration failure.

E. Pleiotropic Effects of Inflammation

Inflammation is pleiotropic in that while both resident and infiltrating immune cells help potentiate their destructive phenotype soon after injury, modulating this early response in a proper way may have beneficial effects (311). Thus selectively blocking proinflammatory factors such as NFκB (40) or curtailing, but not totally eliminating, monocyte infiltration into the cord following injury does help to preserve more myelinated axons, reduce cavitation, and improve functional locomotor activity (145, 310). Specific inhibition of neutrophil entry into the lesion through antibodies targeting rolling adhesion or by other means has also been found to be neuroprotective and beneficial following SCI (19, 137, 381), possibly due to the decrease in proinflammatory factor secretion (372, 396). However, completely abolishing the CNS immune response following injury seems to worsen functional outcome as well as expand the lesion area. For example, genetic ablation of the alarmin IL-33 resulted in the long-lasting failure of monocyte infiltration following SCI but, surprisingly, an eventual decrease in neuronal survival and functional recovery compared with wild-type controls (127). Globally inhibiting inflammation using high doses of glucocorticoids has yielded conflicting results (72, 108, 155) and in some cases increased animal death following SCI (249). To be clear, the medical use of glucocorticoids such as methylprednisolone is no longer recommended in human cases of SCI due to significant complications (339). The wound-healing properties of the late inflammatory response may be one explanation of how complete ablation of inflammation following SCI worsens functional outcome (104, 127, 249).

Inflammation may be modulated in other ways to be beneficial in hastening wound healing and even promoting neurite outgrowth in certain models of CNS injury (37, 288). The inflammatory post-SCI environment, for example, can be modulated by systemic, trophic factor producing stem cells (13) such as multipotential adult progenitor cell treatment (FIGURE 3E) to confer neuroprotection (89, 309, 385, 404) [possibly via their effects in the spleen (88)] and to promote neurite outgrowth despite a growth inhibitory scar barrier (47, 136, 201). Intraspinal injections of anti-inflammatory cytokines such as IL-4 have also been reported to increase M2 or “resolution-phase” macrophages and hasten wound healing (122).

The M1 and M2 paradigm has evolved (69, 318) as more recent studies have begun to stress that these two phenotypes are a continuum instead of a binary (69). Further complicating this story is the innate heterogeneity of both normal microglia and macrophages as well as the heterogeneity of their responses to injury (82). Generally, much of the detrimental effects seen from macrophage interactions following injury have been attributed to their M1-like state. It is M1 macrophages that have been reported to orchestrate glutamate and nitric oxide-induced neuronal death (292, 430). However, through modulation of the processes that convert classically reactive M1 to M2-like macrophages, axonal dieback decreases (47, 89, 426) perhaps through reduction of gliosis and subsequent diminished expression of proinflammatory cytokines such as IL-1β (426). Recently, Kroner et al. (212) have presented a possible explanation as to how this M2 phenotype may be normally repressed. Interestingly, macrophage phenotype (whether they are closer to M1 or M2) seems to be influenced by the type of debris they internalize. Phagocytosis of myelin debris, for example, promoted an M2 bias, whereas phagocytosis of iron, such as that found in red blood cells hemorrhaged into the lesion, pushed classical activation of macrophages towards an M1 bias. This occurred in a TNF-dependent manner which ultimately created more ROS and increased apoptosis in the injury environment.

Taken together, we now appreciate a more nuanced view of inflammation following SCI where although full-forced ablation of the inflammatory response may reduce wound healing and worsen outcome, modulation of the inflammatory milieu towards a “regeneration-associated” phenotype may encourage more favorable outcomes (217).

III. MAJOR CELL TYPES IN THE GLIAL SCAR

With the influx of inflammatory cells, resident glia including astrocytes and oligodendrocyte progenitor cells of the spinal cord become activated and undergo a set of shared characteristics (FIGURES 4 and 6). These include retraction of normally multi-branched processes, enhanced local proliferation in the lesion penumbra (FIGURE 4) (272, 324, 423), and in most cases migration to the lesion margin where they secrete proinflammatory factors. While, over time, the scarring process helps to resolve inflammation, persistent low-grade inflammation by lesion core macrophages and subsequent gliosis remodels the ECM with an increase of fibronectin, collagen, and laminin in the lesion center (109). This substructure is often referred to as the fibrotic component of the scar which becomes replete with pericytes/fibroblasts and macrophages but minimal in microglia (105). The glial component of the scar consists of reactive astrocytes, NG2⁺ oligodendrocyte precursors, and microglia in the penumbra. Additionally, a gradient of lectican-family CSPGs radiates from the lesion center that creates an inhibitory environment for axons chronically (FIGURE 7) (85). While each cell type is discussed discretely, they interact with each other culminating in a permanently remodeled tissue. We will characterize their reactivity following inflammation, how they contribute to the formation of the glial scar, and how this may influence the scar’s axon growth prohibitive effects.

FIGURE 4.

Increased cell proliferation following spinal cord injury. Inflammation following spinal cord injury drives proliferation of many cell types. A–C: quantification of fate-mapped cells in hemisected mouse spinal cords including ependymal (FoxJ1-CreER), astrocyte (Cx30-CreER), and oligodendrocytes progenitor cells (Olig2-CreER) shows proliferation 2 wk and 4 mo following injury. In addition to proliferating, some cell types differentiate including ependymal cells, which differentiate into astrocytes and mature oligodendrocytes. [From Barnabé-Heider et al. (21), with permission from Elsevier.]

FIGURE 6.

The glial scar is composed of multiple cell types. Cells become activated, proliferate, and together form the glial scar as seen in the following sagittal mouse sections. A: fibroblasts visualized using Col1alpha1 promoter at 3, 7, and 14 days following mouse SCI originate from blood vessels and proliferate to form fibrotic component of glial scar. B: astrocytes (GFAP, white) and hematogenous macrophages (tdTomato driven by lysM promoter, red) at 5, 7, and 14 days following mouse contusive SCI. [A and B from Zhu et al. (444), with permission from Elsevier.] C: NG2 staining of oligodendrocyte progenitor cells in mouse dorsal column crush at 7, 14, and 21 days. [From Filous et al. (113), with permission from Society for Neuroscience conveyed through Copyright Clearance Center, Inc.] D: cartoon depicted progressive glial scar formation. Activated and proliferating fibroblasts and macrophages occupy the lesion core of the glial scar by around 14 days post SCI. Activated astrocytes and NG2+ oligodendrocytes occupy the lesion penumbra.

FIGURE 7.

The axon-inhibitory mature glial scar. The mature glial scar becomes a chronically axon-inhibitory structure. A: transplanted dorsal root ganglion neurons (green) are stalled by the glial scar as visualized by GFAP (red) with a gradient of CSPG (CS-56, blue) in rat. [From Davies et al. (85), with permission from Society for Neuroscience conveyed through Copyright Clearance Center, Inc.] B: dystrophic growth cones (arrows and inset, TuJ1, white) are stalled by the glial scar as long as 42 yr in a human case of spinal cord injury. [From Ruschel et al. (330), with permission from AAAS.] C: sagittal section of rat spinal cord 56 days after contusive injury with GFAP (red) and PDGFR-β (green) depicting fibroblasts within the lesion core. [From Zhu et al. (445). Copyright Mary Ann Liebert, Inc.] D: cartoon of mature glial scar depicts reactive astrocytes including palisading astrocytes and NG2+ oligodendrocytes at the lesion penumbra. The lesion core is occupied by macrophages and fibroblasts. Axons become dystrophic as they approach the gradient of CSPGs.

A. Astrocytes and the Glial Scar

Astrocytes are now considered an active player in CNS function contributing to blood-brain barrier maintenance, synaptic physiology, trophic and metabolic support, and a myriad of other homeostatic mechanisms (204). In this section, we briefly review the hallmarks of astrocyte reactivity to the inflammatory environment following SCI and how this response may differ based on a growing appreciation of astrocyte heterogeneity and plasticity. We then describe the astrocytes’ necessary role in acute wound healing and how this tissue remodeling eventually produces a chronic inhibitory structure to axon outgrowth. It is important to emphasize that like the pleiotropic nature of inflammation, astrocytes and the glial scar in general pose a double-edged sword: while they are vital for the acute containment of inflammation, and thus prevention of the expansion of inflammatory processes that can cause greater necrotic damage to cells, reactive scar-forming astrocytes ultimately contribute to the chronic failure of axon regeneration.

1. Astrocytes: a heterogeneous population defined by environmental niches

An area of intense interest is whether phenotypic variances between astrocyte subpopulations could contribute to differences in their ability to hinder or potentiate axon regeneration. Different regions of the CNS can be defined by the heterogeneity of their resident astrocytes (102). Recently, Farmer et al. (106) demonstrated that neurons themselves are capable of diversifying astrocytes through Shh signaling to better contour astrocyte function to different subpopulations of neurons in the cerebellar cortex. Astrocytes are bound to specific regions set after development and while they proliferate at the inner lesion margin after injury, long-distance migration of these newly proliferated astrocytes well beyond their specified regions does not occur (398). The site of injury to the spinal cord, therefore, plays a role in specifying differential astrocytic responses to injury as well as axon regeneration. For instance, protoplasmic astrocytes have been observed to proliferate less than fibrous astrocytes following SCI (103), while juxtavascular astrocytes have been noted to be the source of the majority of newly proliferated astrocytes (20). In another example, Aldh1l1-positive astrocytes prevalent in the spinal cord have been shown to be less conducive to supporting synaptogenesis compared with their brain-derived analogs (186). Additionally, β-amyloid has been reported to activate astrocytes derived from the cortex and hippocampus into a scarlike state, but not those harvested from areas that tend to lack plaque formation such as the spinal cord or cerebellum (174). In fact, astrocytes cocultured with neurons displayed differential effects on neurite branching and length depending on whether they were derived from the medial or lateral sectors of the midbrain (133). Radial astrocytes (tanycytes) located within the median eminence of the hypothalamus do not scar and allow for regeneration of neurohypophysial oxytocinergic and vasopressinergic axons even after penetrating lesions (68). How these differences in environmental glial niches translate to functional benefits following injury in other regions will require further investigation.

2. Reactive astrocytes: heterogeneous response to inflammation

Following exposure to the inflammatory environment post injury, astrocytes become reactive by a process often described incorrectly as astrogliosis (i.e, more astrocytes). Indeed, much of the astroglial response to injury is due to cellular hypertrophy and morphological rearrangements (i.e., reactive astrocytosis) rather than proliferation. Generally, astrocytosis has most commonly been marked by varying upregulation of the intermediate filament decorating proteins glial fibrillary acidic protein (GFAP) and vimentin (422), hypertrophy of the primary branches (373, 412), expansion of normally defined astrocytic domains at the site of injury (408), and very restricted proliferation adjacent to the lesion margin (20, 408). The increase in GFAP, in particular, helps to allow scar astrocytes to form a rigid, densely bundled structure around the lesion (FIGURE 5A) (421).

FIGURE 5.

Astrocyte plasticity following spinal cord injury. Astrocytes are a heterogeneous population of glia that are highly plastic and whose phenotype also depends on environmental factors. A: astrocytes (GFAP, green) activated by the post-injury inflamed environment (14 days after mouse crush SCI) become “wall-like” to wall-off fibroblasts (fibronectin, red) of the lesion core (LC) at the astrocyte scar border (ASB). Astrocytes prevented from becoming activated through STAT3-KO, however, become more “bridge-like” instead at the scar border. [From Wanner et al. (408), with permission from Society for Neuroscience conveyed through Copyright Clearance Center, Inc.] B: astrocytes (GFAP, green) in AAV-shPTEN mice create bridges for BDA-labeled cortical spinal tract axons (red) to cross the lesion by 8 wk following crush injury. [From Zukor et al. (447), with permission from Society for Neuroscience conveyed through Copyright Clearance Center, Inc.] C: astrocytes grown in vitro (GFP, green) and transplanted into naive, 7, or 14 days post spinal cord contused mice adopt inflamed “wall-like” or axon-conducive “bridge-like” phenotypes based on their environments as visualized through surrounding astrocyte (GFAP, red) staining. [From Hara et al. (157), with permission from Macmillan Publishers Ltd.]

Astrocytosis occurs as a direct response to physical trauma, although a septic inflammation even without direct physical injury has been shown to induce strong astroglial reactivity (115). For example, a tiny microinjection of zymosan (a potent yeast cell wall inflammogen) into the white matter, while minimally physically destructive, nonetheless was sufficient to locally activate macrophages and cause robust GFAP and CSPG upregulation and rapid migration of astrocytes outwards from the injection site to densely encircle the zone of inflammation (115). Reactive astrocytes in the vicinity of the SCI-induced inflammatory storm respond through intramolecular signaling changes to the myriad of secreted cytokines, alarmins, and other cellular signals. Enhanced calcium signaling in astrocytes as a response to the inflammatory storm, for instance, activates them (191) to promote proreactive transcriptional changes (130). Proinflammatory cytokines, notably TGF-β and fibrinogen infiltrating from the periphery through a compromised BSCB, have been shown to activate SMAD signaling through p75 cleavage and subsequent nuclear pore complex remodeling (327) to enhance GFAP expression (407). The STAT signaling pathway has also been implicated in astrocytosis to initiate hypertrophy (FIGURE 5A) (165, 291, 408). Many of these cellular signaling pathways converge to enhance transcriptional activation and subsequent expression of activation-type proteins such as GFAP (262).

The extent of reactivity and strength of the cellular response and subsequent transcriptional changes are dependent on the type of lesion, regional heterogeneity of the astrocytes themselves, and distance from the injury. Different types of injuries elicit different transcriptional changes in astrocytes (436). Zamanian et al. (436) have found that 50% of the genes upregulated in reactive astrocytes were different between lipopolysaccharide (LPS)-evoked and ischemic injuries. Recently, Liddelow et al. (236) have pinpointed the necessity of microglia to activate a subset of astrocytes through TNF, IL-1, and complement component 1 subcomponent q signaling following LPS administration. In fact, Csf1r knockout mice, which lack microglia, fail to activate this subtype of astrocytes (236). This subset of reactive astrocytes has been reported to upregulate a specific group of genes in response to such proinflammatory signaling and have been found to be potently neurotoxic and detrimental to oligodendrocytes (236). Other microarray or single-cell gene expression studies have confirmed the transcriptional heterogeneity among reactive astrocytes (7, 15, 157, 186, 331). Even in primary astrocyte cultures, single-cell gene expression analyses have revealed transcripts of vimentin and GFAP ranging from 50,000 to <100 which stresses that astrocyte heterogeneity is more complex than a binary expression pattern of proreactive markers (366). Most of these transcriptional changes were found to be transient following injury and resolved around 1 wk as inflammation subsided (79, 161, 320). Additionally, the type of injury, severity of trauma, and the distance of astrocytes away from the lesion core are all critical in eliciting a range of responses (431, 439). Thus the farther away the astrocyte is from the lesion epicenter, even as axons and oligodendrocytes are dying in their midst, the less severe the reactivity (289, 408). Importantly, astrocytes close to the lesion penumbra change their orientation to form an overlapping wall of densely packed and adhered cells that lack fine ramifications (408). This is in contrast to astrocytes farther away from the lesion which display a lessening gradation of hypertrophy including decreased thickening of processes and minimal overlapping of domains. Ultimately, cell signaling and transcriptional changes following the inflammatory storm enable potentiation of astrocytosis including cyotoskeletal protein expression (262), proliferation (21, 22, 408), and secretion of proinflammatory factors (175) to allow for an adequate wall building response to insult.

Reactive astrogliosis (i.e., astrocyte proliferation) is typically confined to the inner margin of the penumbral territory directly adjacent to the inflammatory core (FIGURE 6B). Following injury, reactive astrocytes have been known to derive from ependymal cells (261), NG2⁺ oligodendrocyte progenitors (443), but mostly from other reactive astrocytes (142). Thus an important question in considering astrocyte heterogeneity is the parental source of newly proliferated astrocytes and where these proliferating niches occur. Following a dorsal spinal cord cut injury, Barnabé-Heider et al. (21) observed that ependymal cells near the center of the lesion gave rise to astrocytes located closest to the inflammatory core, while astrocytes slightly farther into the penumbra of the lesion self-proliferated. However, the critical role of ependymal cells in scar formation has recently been called into question since their contribution appears to be meager at best and seems to be dependent on the need for a direct hit on the cells determined by the precise location and severity of the SCI lesion (321).

Another hallmark of reactive astrocyte wall building is the increased deposition of ECM proteins, notably axon growth inhibitory CSPGs seen in a gradient pattern around the lesion in vivo (84). This aspect of reactive astrocytes will be detailed in another section.

3. Reactive astrocytes: differences in immature and mature astrocytic responses to injury

The maturation state of the astrocyte is another major contributing factor to the heterogeneity of its response following injury (27, 168, 278). As an example, immature astrocytes mount a weak reactive response to proteins such as β-amyloid (58), whereas mature astrocytes become highly reactive to the same substrate (329). This difference in reactivity may be one explanation of how immature astrocytes promote axon regeneration better than mature astrocytes (111). Thus astrocytes cultured to maturity for 35 days or longer in vitro do not themselves readily extend past a high concentration of CSPGs, whereas immature astrocytes cultured for less than 2 wk can secrete matrix-degrading enzymes that allow them to cross this inhibitory terrain (111). Differences between immature and mature astrocytes are also seen in vivo in their ability to wound heal. Indeed, transplantation studies have shown that immature astrocytes are far more reparative than mature astrocytes (83, 153, 418). Mature astrocytes transplanted into the brain appear to stimulate increased macrophage and fibroblast entry and cavitation (111). In contrast, immature astrocytes derived from neonatal cortices carefully expelled along the track of a retracting needle within the cingulum bundle (111) or precoated onto nitrocellulose scaffolds and impelled into the adult forebrain have been shown to “knit” together the tissue bordering the lesion as well as reduce scar formation associated with the implant (153, 189, 355). Importantly, axon growth is enhanced when immature astrocytes are present (28, 111, 152) with minimal astroglial scarring (329). Tanycytes of the median eminence may represent a more primitive form of astrocyte which allows for regeneration after injury (67). Intriguingly, in the naked mole rat, reactive astrocytes allow for regeneration in the optic nerve well beyond a crush lesion (302). One may speculate that naked mole rat astrocytes remain in an immature state to allow for axon regeneration much like the primitive radial glia of highly regenerative zebrafish (244). Thus the overall agedness of the CNS environment and the maturational state of astrocytes play a role in differential astrocyte functions.

4. Reactive astrocytes: bridge building or wall building?

In the adult, reactive astrocytes play a multifaceted, complicated role in SCI. Usually, the level of astrocytosis following typical trauma to the spinal cord is severe enough to elicit permanent tissue remodeling and wall building. Histological observations of the glial scar show that the general morphology of astrocytes after typical severe lesions may not be conducive for axon outgrowth. Namely, border-forming astrocytes at the innermost edge of the lesion penumbra create palisading-like patterns with thick hypertrophied processes that densely overlap and pack around the lesion (408). Border-forming astrocytes additionally produce potently inhibitory ECMs in response to a variety of different cell and molecular triggers (57, 157, 336). This three-dimensional, wall-like structure may pose a physical and chemical barrier to axon outgrowth. Interestingly however, the upper surfaces of intensely reactive astrocytes themselves, grown in vitro as a monolayer upon β-amyloid, a potent inducer of reactive astrocytosis, can be conducive to neuronal outgrowth because they position their inhibitory CSPG-laden ECM only on their undersurface abutting the amyloid substrate (57, 342). Astrocytes are, therefore, highly plastic and dynamic depending on their polarity, age, lineage, and extent of the inflammatory environment that they encircle.

To further probe the role of astrocytosis following injury, several groups have utilized transgenic mice to prevent astrocytes from becoming reactive. When STAT3 signaling was specifically perturbed in astrocytes following injury, astrocytosis was decreased with a lack of hypertrophy and reduced GFAP expression compared with wild-type astrocytes (FIGURE 5A) (165, 291). The perturbation of post-injury astrocytosis signaling also prevented the formation of palisading and densely packed wall-building astrocytes at the penumbra (FIGURE 5A) (408). What followed after early disruption of the formation of the glial scar wall included an expansion of the lesion site with increased inflammatory cell infiltration resulting in further loss of function. In studies where GFAP-expressing astrocytes were specifically targeted for cell death, leaky BSCB persisted (107) and inflammatory cells from the periphery increased up to 25-fold (50). As a result, neuronal degeneration increased, as did edema. Moreover, wound repair following stab or crush injury was incomplete as the astrocytic scar border failed to close (50, 107). Ultimately, the interruption of post-injury glial scar formation from the beginning, leading to BSCB related complications, worsened functional recovery through an unchecked inflammatory response. Clearly, completely removing astrocytes in an immediate post-injury environment worsens functional outcomes as seen recently in Anderson et al. (6) (and further discussed below). However, aside from glial scar formation, certain subtypes of reactive astrocytes have a myriad of net positive neuroprotective effects following injury such as expression of key proteins to facilitate stabilization of the chaotic milieu. These include limiting edema through an increase in aquaporin channel 4 expression, limiting neuroexitotoxicity through glutamate transporter upregulation, reducing oxidative stress through glutathione production, and increasing trophic and metabolic support for compromised neurons (112, 362).

But what happens to astrocytes and how do they respond when far less disastrous perturbations occur? Zukor et al. (447) have demonstrated how differences in lesion environments affect reactive mature astrocytes and result in either positive or negative consequences to regeneration. For example, when corticospinal tract axons are strongly growth enhanced through PTEN knockdown after a very thin (<0.5 mm) lesion is created in the spinal cord with limited fibroblast and macrophage infiltration, reactive adult astrocytes could form narrow bridges upon which growth-enhanced axons were able to cross and bypass the lesion (FIGURE 5B). This is in contrast to more expansive lesions that were associated with large numbers of fibroblasts/pericytes and macrophages. The increase in inflammatory and mesenchymally derived cells promoted the construction of a wall around the lesion resulting in regeneration failure. In other situations, when GFAP and vimentin are conditionally knocked out before injury, astrocyte reactivity including stereotyped hypertrophy and proliferation is reduced (262). In paradigms like this where astrocytes are not killed but, rather, prevented from becoming severely hypertrophied and rigid (422), some axon outgrowth and functional improvements can occur (157, 262).

Other strategies involving mild astrocyte manipulations have resulted in increased axonal growth (177, 179, 246). For instance, application of fibroblast growth factor (FGF), which has been previously shown to increase astrocyte plasticity towards a more bipolar, immature phenotype (192), aided axon growth into a peripheral nerve graft (88). Axons have been observed to enter the region of the peripheral nerve graft where reactive astrocytes are aligned and inhibitory proteoglycan matrices are degraded with chondroitinase ABC (ChABC) (216, 254) as opposed to regions containing perpendicularly oriented, wall-like reactive astrocytes (408). In another example, auditory neuroblasts laid superficially along reactive astrocytes of the crushed CNS portion of the eighth nerve, but only with the addition of ChABC, were able to invade the reactive glial environment which acted as a bridge that supported neurite elongation and functional reinnervation (342). Importantly, recent work from the Okada group (157) has elegantly illustrated the plastic nature of astrocytes that better contextualizes the results in Anderson et al. (6), which we believe erroneously interprets that the astrocytic scar always aids axonal regeneration (FIGURE 5C). In Hara et al. (157), reactive astrocytes genetically labeled with GFP under the Nes promoter were FACs sorted from 7-day post-injured spinal cords and grafted into a new cohort of mice that were either intact or injured. Further sequencing 7 days after the graft showed that these astrocytes conformed to the environment of the host, forming scar astrocytes in injured animals or reverting to an unreactive, quiescent state in the naive cord (157). Importantly, astrocyte reprogramming in an injured environment seems to be instigated, at least in part, by fibrotic extracellular matrix material in the lesion core. In particular, type I collagen that is highly secreted following SCI (95) acts upon astrocytes through integrin receptors leading to subsequent N-cadherin signaling to form a tight scar (157). Interfering with specific integrin signaling only during the later scarring phase of astrocytosis after SCI with the use of antibodies reduced the scar (but did not kill the astrocytes) and allowed for significant axonal regeneration around the edges and beyond the lesion. This is in contrast to the overly simplistic interpretation by Anderson et al. (6) that astrocytic scars are wholly beneficial in regeneration because their techniques to kill or misalign scar-forming astrocytes immediately after injury did not lead to spontaneous regeneration. Anderson et al. (6) used transgenic STAT3 knockout mice or mice with ganciclovir-targeted astrocytes under the GFAP promoter. Conditions like this which globally affect astrocytes acutely after injury, in turn, unleash the core of inflammatory macrophages that are highly toxic to regenerating axons (see discussion above and Refs. 50, 231). We take a more nuanced view that astrocytes are highly plastic cells and their ultimate phenotype, either wall building or bridge building, depends on the intensity of the inflammatory environment they occupy and the growth capacity of axons in their vicinity (351).

The view of astrocyte biology following SCI and the appreciation of their polarity and plasticity especially in the presence of very small lesions or where core inflammation and pericyte/fibroblast proliferation are at a minimum have evolved dramatically (77, 350, 351). Astrocytes are a heterogeneous population that conform to different environmental niches and interact extensively with other cell types to potentiate the density of the glial scar. In the midst of very minimal lesions where collagen-producing fibroblasts/pericytes and inflammatory cells are lacking, but also in the presence of robustly growing axons (162, 184), astrocytes will abandon their wall-building role to favor bridge building and can be supportive to regeneration especially when their inhibitory matrices are reduced. However, their reactivity in the vicinity of typical large lesions filled with fibroblastic and immune cells favors wall building, which hinders axon regeneration. Nonetheless, astroglial scar barriers are not absolute and will allow the passage of some regenerating axons if they are maximally growth stimulated (240–242). Like macrophages, we may be able to take advantage of this astrocytic pleiotropy and modulate their reactivity to improve functional outcomes (411).

B. Oligodendrocytes and Their Progenitors

Oligodendrocytes proceed through sequential stages as they differentiate, including a pre-progenitor state and an immature state where the cells transiently express the NG2 CSPG before possessing the ability to myelinate in their mature form (26). In the adult cord, satellite oligodendrocyte progenitors of the grey matter and myelinating oligodendrocyte progenitors of the white matter constantly proliferate to sustain a coterie of progenitor cells as well as to generate increasing numbers of mature oligodendrocytes (21). Oligodendrocyte progenitor cells (OPCs) expressing the purportedly inhibitory NG2 proteoglycan, collectively called polydendrocytes, are one subpopulation that has been the most widely studied in models of SCI since their discovery in the 1980s (233, 367). They are characterized by their branched processes and ability to normally self-renew in adulthood (172). However, other cell types also express the NG2 CSPG. These include pericytes (346), reactive Schwann cells (259), and macrophages (188, 271); thus immature NG2⁺ oligodendrocyte precursors may be more precisely identified in conjunction with other oligodendrocyte markers (26).

Physical impact to the spinal cord causes mature and immature (i.e., precursor cell) oligodendrocyte apoptosis at the lesion site (146, 237). Oligodendrocyte depletion due to secondary injury may continue for weeks rostral and caudal to the lesion along degenerating axon tracts (234) perhaps as a result of the loss of axon-derived trophic factors. NG2⁺ OPCs proliferate following injury peaking at 5 days (FIGURE 6C) (245, 435) to accumulate in the lesion epicenter and penumbra (245, 260). This was shown in a model of incomplete spinal cord transection by Barnabé-Heider et al. (21) who used genetic fate mapping in combination with BrdU colabeling to reveal that the rate of oligodendrocyte progenitor proliferation had doubled (FIGURE 4). NG2⁺ OPCs may also derive from other glial sources. Genetic fate mapping further revealed that a small subset of ependymal cells differentiate into NG2⁺ oligodendrocytes (21). This phenomenon has been confirmed by other studies (170, 261).

Secondary injury, in part due to increased cytokine levels such as TNF-α from other activated glia, causes OPCs to become reactive resulting in increased expression of the purportedly inhibitory NG2 CSPG. Their increased rates of proliferation (352) owe in part to noncanonical STAT3/SOCS3 signaling (154) and decreased differentiation into mature oligodendrocytes (322). While a few NG2⁺ OPCs have been shown to differentiate into reactive astrocytes following CNS injury, they are not the main source of de novo astrocytes (208). Some Schwann cells occupying the lesion site (41, 45) have also been shown to differentiate from NG2⁺ OPCs (437). Activated NG2⁺ oligodendrocytes also show a change in morphology including withdrawal of long, complex branches into shortened hypertrophied processes concomitant with swelling of the soma and accumulation into dense plaques surrounding the lesion site as a result of enhanced β-catenin signaling following injury (325). NG2⁺ glia additionally contribute to the inflammatory milieu (215) by activating macrophages and astrocytes (325) and secreting proteases such as MMP-9 (343) that help increase the permeability of the BSCB (232, 285).

What is the impact of these newly proliferated, CSPG-producing NG2⁺ oligodendrocyte progenitors on axon regeneration in the lesion penumbra? Importantly, these newly proliferated, and supposedly axon inhibitory, NG2⁺ cells have been shown to highly colocalize with dystrophic axonal growth cones (187, 259, 284). Whether NG2⁺ OPCs themselves inhibit axon regeneration has been a matter of considerable contention (232, 429). Earlier studies had described how the NG2 CSPG itself inhibits axon regeneration in vitro (98, 359) and in vivo and that this effect could be reversed using an NG2-blocking antibody (245, 378). However, recent studies have revealed that the NG2⁺ OPC itself can first entrap then stabilize retracting growth cones after a dorsal column lesion despite the extensive presence of macrophages which cause the dying back phenomenon (48). In addition to the proteoglycan NG2, these cells express an abundance of fibronectin and laminin that is likely critical for balancing the ECM and promoting the entrapping rather than repulsive effect on axons. It is also possible that pericytes (which also express NG2, laminin, and fibronectin) may play a role in entrapping dystrophic axons at least transiently within the lesion core.

Closer inspection of the NG2⁺ OPC and dystrophic growth cone interface has revealed synaptic-like structures by which axon tips may become further stabilized along the lesion penumbra (113, 363). NG2⁺ cells normally receive excitatory axonal synaptic-like contacts in the developing and adult brain that may likely play a signaling role in allowing the OPC to ‟sense” if a demyelinating event has occurred in its vicinity (32, 128, 446). Following SCI, the abundance of NG2⁺ OPCs in the lesion penumbra may be providing a safe haven in the midst of fulminant inflammation that serves to stabilize the severed axonal ending through this unusual glia/neuron bond and, unfortunately, helping to prevent them from regenerating further (48, 112, 125). Synaptic-like contacts between regenerating sensory axons that abruptly halt their forward progress upon NG2⁺ glia has also been documented at the dorsal root entry zone after root crush (363). How long this entrapment lasts may depend on whether the oligodendrocyte progenitor eventually changes its phenotype to become less hospitable. Whether axons eventually dieback to their sustaining collaterals as Cajal had predicted if and when the growth cone escapes this entrapment remains an unanswered question. However, dystrophic growth cones have been observed to persist in the spinal cord scar for 40 yr in a human case of SCI, suggesting that some severed axonal tips can persevere in a doomed, synaptic-like state with glial cells indefinitely (FIGURE 7B) (330). Finding a means to free the dystrophic growth cone from its synaptic entrapment may prove to be therapeutic (384).

Oligodendrocyte progenitors themselves are affected by the CSPG-content of the glial scar which can spread rostral and caudal to the lesion following SCI (FIGURE 6C) (8). While NG2⁺ OPCs express the NG2 CSPG, other CSPGs of the astroglial scar such as versican and neurocan have been shown to potently inhibit oligodendrocyte myelination (307, 317) which may be contributing to chronic remyelination failure following SCI (34, 46, 100, 394) as well as multiple sclerosis (198). OPCs express the receptor protein tyrosine phosphatase sigma (RPTPσ and see below for further discussion of CSPG receptors), which in the presence of CSPGs inhibit vital oligodendrocyte functions such as proliferation, differentiation, migration, and myelination through ROCK and Rho downstream signaling (307). Oligodendrocyte progenitors in the presence of CSPGs show decreased outgrowth of processes and differentiation into mature, myelinating oligodendrocytes (48, 194, 348). Oligodendrocyte migration is additionally compromised by the sugar moieties of CSPGs, which can be reversed by the sugar-cleaving ability of chondroitinase ABC (194, 349). Changes in Akt and ERK signaling through RPTPσ and CSPG binding also decrease oligodendrocyte survival and maturation (100). NG2⁺ cells surrounding the lesion penumbra, despite possessing the ability to robustly proliferate following injury, display reduced ability to differentiate to ultimately resume their necessary function of remyelinating denuded axons chronically (377, 394), although disperse remyelination has been shown to occur (166). The proximity of OPCs to the CSPG-rich scar and subsequent CSPG signaling may be one explanation of how oligodendrocyte function is compromised following injury. Thus the CSPG-rich content of the lesion contributes to an environment that serves to curtail remyelination as well as regeneration.

C. Pericytes, Fibroblasts, and the Fibrotic Component of the Glial Scar

Fibroblasts are the major connective tissue cells found throughout the body. These cells provide a structural framework partly through deposition of ECM components, and only invade or are produced in the CNS after injury (1, 109). Following a stab injury, for example, meningeal or perivascular fibroblasts migrate to the lesion where they proliferate and contribute to the development of the fibrotic scar in the lesion epicenter (FIGURE 7C) (59, 109). Fibroblasts aid in the contraction of the lesion and subsequent wound closure, and although the fibrotic scar is essential to the acute healing process after injury, it also contributes a portion of the chronic impediment of the glial scar to axon regeneration (401). Fibroblasts have been shown to decrease neurite length in vitro (286, 329) as well as in vivo (401, 444). Even robustly regenerating PTEN knockdown axons have been reported to avoid the fibroblast-filled lesion epicenter (447). This phenomenon may be due in part to the secretion and deposition of axon-repulsive cues by fibroblasts (303) such as tenascin, versican, and collagen. Recall that type I collagen is a major trigger of astrocytic scar formation (157).

Pericytes have recently been found to contribute to the fibrotic scar following CNS injury by chronically differentiating into fibroblast-like cells (142, 361). Normally, pericytes envelope the blood vessels of the CNS to aid in the control of blood flow, support angiogenesis, help maintain the blood-brain barrier, and perform other homeostatic functions (413). However, following injury and disruption of the BSCB, a particular subset of pericytes delaminate from the basal laminae of the blood vessels they envelope and migrate to the lesion core (142). In a model of spinal cord dorsal hemisection, Görtiz et al. (142) found that the lesion site becomes devoid of blood vessels ~1 day following injury, but by 5 days after injury, blood vessels had regenerated and pericytes had continuously proliferated. This perivascular source of proliferating fibroblasts has also been observed in traumatic brain injuries (97) as well as contusive SCI where the dura mater remains intact (361). However, recent work by Guimaraes-Camboa et al. (149) (through the use of fate mapping with the help of a very specific pericyte marker, tbx18) has posited that pericytes may not be the most prevalent fibrogenic progenitors. However, in support of the fibrogenic pericyte hypothesis, Soderblom et al. (361) genetically labeled collagen1α1-producing cells and showed that this population greatly proliferated to contribute to the fibrotic scar. Like the cell types described in Görtiz et al. (142), Soderblom et al. (361) found that they are also PDGFR-β and CD13 positive. According to these two studies, pericytes at the lesion core lose expression of pericyte-markers, including CD13 and PDGFRα, and upregulate fibrogenic markers such as fibronectin (142). Whether these two mouse studies have independently identified the same source of fibrotic scar formation will need to be clarified along with elucidation of the exact cellular identity of these CD13/PDGFRα+ cells. It is also possible that pericytes (which also express NG2, laminin, and fibronectin) may play a role in entrapping dystrophic axons at least transiently within the lesion core.

Recent studies in mouse models of contusive SCI have additionally shown that the increase of fibroblasts at the lesion site correlates with the infiltration of leukocytes at 7 days post injury where they migrate and concentrate in the lesion core by 14 days (FIGURE 6A) (444, 445). Zhu et al. (444) have further suggested that the delaminated fibroblasts are called to the lesion site through infiltrating macrophages instead of resident microglia as clondronate-induced depletion of macrophage entry greatly decreased formation of the fibrotic component of the scar. This migration of fibroblasts into the cord as well as their activation may be due to TNF and BMP secretion by macrophages (444). TGF-β1 is another potent proinflammatory cytokine expressed in the lesion that has been shown to induce reactive fibrosis in culture (205) which, in turn, has been shown to induce reactivity in other glia such as astrocytes (409). Maturation of the fibrotic scar includes deposition of a basal lamina, formation of the glial limitans structure to sequester fibroblasts from the rest of the CNS (44), and BSCB reformation (255). Basal lamina formation involves cell-cell contact of EphB2 receptors found on fibroblasts and Ephrin-B2 receptors on reactive astrocytes and subsequent secretion of ECM molecules (205). Recent studies involving specific deletion of fibronectin in macrophages have shown that fibroblasts are, indeed, the cell type responsible for the majority of fibronectin secreted in the ECM (445). Soluble fibronectin is possibly integrated directly into the ECM via activated macrophages through integrin receptor activity (445). Overall, this process of fibrotic scar formation is essential to the wound healing process of the lesion after injury. Indeed, genetic, large-scale ablation of pericytes resulted in the enlargement of the lesion and failure of the lesion to contract and close, but discussion was lacking as to whether there was enhanced regeneration (142). Thus it will be important in the future to learn whether specific ablation of pericytes can reduce the fibrotic component of scarring and also stimulate regeneration.

Given the increasing breadth of knowledge regarding fibroblasts or fibroblast-like cells and their interactions with various glial cell types in the forming scar, it is imperative that we consider that no single cell type exists in isolation following SCI. Indeed, glial and fibrotic scar formation is an intricately synchronized and interactive cell process with many potential therapeutic targets for functional recovery.

IV. CHONDROITIN SULFATE PROTEOGLYCANS AND THEIR EFFECTS ON LIMITING REGENERATION/PLASTICITY AFTER SCI

The lesion penumbra surrounding the fibrotic core is marked by reactive glial cells and an abundance of CSPGs (FIGURE 7). While different cells are known to express discrete types of CSPG after SCI, in general, neurocan, versican, brevican, and NG2 predominate at the site of trauma, with expression reaching a plateau 2 wk after the lesion is formed and remaining, albeit at decreased levels, throughout life (11, 12, 51). Interestingly, the axon-inhibitory sugar moieties of CSPGs, glycosaminoglycans or CS-GAGs, are typically sulfated to generate greater amounts of CS-A and CS-C but also CS-E following trauma. The upregulation of CSPGs following SCI is partly caused by the trauma-induced local infusion of blood and fibrinogen which mediates activation of TGF-β (181, 375).

Due to the high CSPG content within the glial scar, it is not surprising that axon regeneration and plasticity are hindered following SCI (FIGURE 8A) (5, 23, 35, 39, 98, 116, 123, 359, 380). The classic demonstration of this inhibitory effect came from Davies et al. (84) who used microtransplantation of adult dorsal root ganglion neurons into degenerating myelin rostral to a lesion of the dorsal columns where, surprisingly, they could regenerate axons robustly among myelin debris as well as reactive intratract astrocytes, demonstrating that reactive astrocytes farther from the lesion can be axon growth permissive. However, once they reached scarred areas just outside of the lesion core that contained wall-building astrocytes and abundant CSPGs, they halted their progress and formed dystrophic endings. Considering the importance of CSPGs to recent research in the SCI field, we will detail the contribution of CSPGs to growth cone dystrophy and discuss their newly discovered cognate receptors (RPTPσ and LAR) as well as putative downstream signaling cascades. Finally, we will end this section with the role of CSPGs in limiting axonal plasticity via the PNN.

FIGURE 8.

Chondroitin sulfate proteoglycans and the growth cone. Chondroitin sulfate proteoglycans (CSPGs) of the glial scar inhibit axon outgrowth through the protein tyrosine phosphatase sigma receptor (PTPσ). A: CSPGs consist of a protein core with varying glycosaminoglycans. CSPGs attach to hyaluronan through linker proteins to form the perineuronal net (PNN) surrounding the soma of select neurons. The glycosaminoglycan segment of CSPGs binds to transmembrane receptor PTPσ to contribute to receptor monomerization causing growth cone dystrophy. Heparan sulfate proteoglycans binding to PTPσ promote their oligomerization to allow for axon growth. B: the PNN is visualized through WFA staining of glycosaminoglycans. [From Massey et al. (251), with permission from Society for Neuroscience conveyed through Copyright Clearance Center, Inc.] C: CSPGs visualized through CS56 antibody staining (green) are upregulated following C2 hemisecton and surround dextran amine Texas Red (DTR)-labeled phrenic motor neurons (PMN). [From Alilain et al. (5), with permission from Macmillan Publishers Ltd.] D: axonal growth cones on polylysine and laminin are visualized through actin (red) and tubulin (green) staining. Dystrophic growth cones are visualized through tubulin and actin staining of adult mouse dorsal root ganglion neurons on aggrecan. (From Hur et al. Proc Natl Acad Sci USA 108: 5057–5062, 2011, with permission from PNAS.)

A. Chondroitin Sulfate Proteoglycans and the Growth Cone

Depending on the geometry of the assay, in response to substrate-bound CSPGs in vitro, growth cones typically either turn or enter a dystrophic state (FIGURE 8D). The repulsive turning effect was initially demonstrated with in vitro stripe assays where neurons would deviate abruptly to avoid growing on CSPG-rich regions (98, 359). The stripe assay promotes outgrowth upon alternating patterns of permissive versus inhibitory substrates to mimic the environment that a neurite might experience at the point of a sharp cellular boundary (359, 360). Through this assay, it has been determined that neurite growth and rate of extension, calcium signaling, growth cone morphology, and filipodial behavior are all significantly affected by the GAG portion of CSPGs (31, 213, 358, 359, 395).

The strength of the stripe assay is its ability to model special in vivo situations such as that which occurs at the glial roof plate barrier that acts in the fashion of a guard-rail to the ascending dorsal columns and descending CST axons during development of the spinal cord (196, 359). Here, an immediate change in substrate, or substrate concentration, dramatically affects neurite growth and guidance to induce turning. However, most often in vivo, changes in substrate concentration occur more gradually in space or time. For example, in the vicinity of the glial scar, neurites attempting to regrow will come into contact with gradually increasing concentrations of CSPGs as they struggle within the lesion epicenter (350). This geometry has been reflected with the development of the spot assay (392). This assay generates a smooth gradient of proteoglycan within a defined area of the spot rim, similar to that which occurs in the glial scar penumbra. Initially this technique was used to show that dorsal root ganglion axons will enter but then stall upon an increasing gradient of aggrecan, eventually resulting in the formation of dystrophic end bulbs (FIGURE 8D) (392). More recently, it has been used to identify CSPG receptor upregulation as well as the critical demonstration that the inhibitory effect of CSPG gradients upon growth cone advance is not due to repulsion but, rather, to an over adhesion of the growth cone to the inhibitory substrate at a particular point within the gradient (221). Such super-adhesion-mediated growth cone inhibition also occurs in relation to NG2+ oligodendrocyte precursor cells which can entrap axonal growth cones via synaptic-like connections (48, 113). It is likely that in vivo growth cone dystrophy is also caused, at least in part, by such overly adhesive interactions with the CSPG GAG chains presented by OPCs (98, 110, 123, 178, 220, 266, 279, 293, 357).

B. Chondroitin Sulfate Proteoglycans Mediate Signaling in the Growth Cone

Another way to overcome the effects of proteoglycans on growth cones would be to interrupt the downstream signaling pathway through which CSPGs mediate their dystrophy producing effect. A number of studies have identified the RhoA/ROCK signaling pathway as being instrumental in causing growth cone collapse, which can precede dystrophy (76, 99), through the prevention of microtubule formation (247). Indeed, use of the specific inhibitors C3 coenzyme or Y-27632 in vivo facilitates axonal growth following SCI (90, 120, 268, 316). Nonetheless, CSPG-mediated growth cone collapse may be further instigated by a number of alternative signaling receptors and pathways. For example, epidermal growth factor receptor (EGFR) is thought to control the calcium signaling within neurites and is inhibited in the presence of CSPGs (78, 210). Furthermore, CSPGs interfere with integrin binding and signaling (2, 296, 379, 448).

The failure of axon growth on CSPGs has recently been attributed to the activation of protein kinase A (PKA) (214). These data are interesting as cAMP, which activates PKA, is known to promote regeneration of optic nerves (216, 269). Indeed, this key intracellular signaling molecule has been shown to increase neuronal regeneration in a number of models (55, 71, 314, 315) through its effect on PKA (131, 239, 335), and by increasing growth factor receptor translocation from the cytoplasm to the cell membrane (263). This may suggest that the effect of PKA activation on neuronal growth is dependent on the specific range of extracellular matrix molecules that the growth cone contacts.

C. Chondroitin Sulfate Proteoglycan Receptor Discovery and Mediation of Growth Inhibition by Excess Adhesion

Less than a decade ago, our best evidence suggested that CSPGs mediate their effect on axon guidance and growth cone dystrophy solely through their negative charge (273) or via interactions that impede growth-promoting factors. Recently, however, a few CSPG receptors have been identified that may directly initiate these inhibitory effects. These are the adhesion-promoting, synaptogenesis-related LAR family receptor phosphatases, RPTPσ and LAR. The third family member, RPTPδ, has not yet been studied in relation to CSPG inhibitory function.

RPTPσ acts as a functional receptor for specific CSPGs (FIGURE 8A) (124, 345). RPTPσ−/− mice demonstrated increased outgrowth of granule and dorsal root ganglion neurons when grown on substrates containing CSPGs (124, 345), while no such effect was shown following growth on inhibitory myelin substrates (345). As such, these receptors have been shown to be of primary importance in the control of neurite outgrowth after CNS trauma where CSPGs are increased. For example, RPTPσ−/− mice demonstrate enhanced regeneration of the injured CST and growth of sensory axons into an SCI lesion site (124, 345). Similar regenerative capacity through knockout of RPTPσ has been shown in peripheral and optic nerve crush models (124, 258, 345, 387, 442). Interestingly, McLean et al. (258) further demonstrated that RPTPσ−/− mice showed faster rates of peripheral nerve regeneration following sciatic nerve crush and bidirectional growth of fibers when nerve transection was combined with immediate repair or allografting. Gardner and Habecker (135) have recently shown that the effect of CSPG-RPTPσ interactions can also effect sympathetic reinnervation following denervation within the forming scar of the heart following ischemia-reperfusion injury. RPTPσ−/− mice showed hyper-reinnervation of proteoglycan-rich, cardiac scar tissue following recovery unlike RPTPσ+/− animals, which displayed no regeneration into the scar as well as marked susceptibility to generate arrhythmias. As such, CS-RPTPσ interactions can be shown to directly affect capacity and rate of regeneration following trauma and scar formation in both the CNS and peripheral nervous system (PNS).

It is believed that four positively charged lysine residues in the first immunoglobulin domain of the RPTPσ receptor mediate the specific interaction with GAGs (345, 432) on both CS and HS molecules (FIGURE 8A) (10, 432). Sulfation patterning additionally affects binding of specific CSPGs to RPTPσ. Due to the nature of electrostatic interactions, the bifunctional RPTPσ is likely to bind most strongly with CS-D and CS-E. Given that the latter of these molecules is highly upregulated after CNS injury (42, 93, 432), this makes the receptor a strong candidate to target in a treatment strategy (42, 221). Importantly, the downstream effectors of the RPTPσ signaling pathway have yet to be identified, meaning that the mechanism through which CSPGs mediate inhibitory signaling is not yet fully elucidated. It is possible that it converges on the Rho/ROCK pathway and aids inhibition of Erk signaling by dephosphorlyation (100, 221).

The second receptor that mediates the cell growth inhibitory function of CSPGs is LAR. Among other things, LAR−/− mice demonstrate developmental deficits within the septohippocampal cholinergic fiber system (207, 428), suggesting an effect in the organization of neural networks, especially through the formation and regulation of synapses (169). LAR binds to itself (428), to HSPGs (121), and to CSPGs (114) (in a dose-dependent manner). Further downstream, the CSPG-LAR interaction has been shown to increase RhoA signaling and inactivate Akt, indicating a convergence of all CSPG mechanisms of action (100, 114).

Following traumatic CNS injury, LAR−/− mice have improved regeneration within the corticospinal tract and serotonergic sprouting leading to some functional recovery of the locomotor system (114, 421). This effect may be a result of the homophilic binding exhibited by the receptor which has been identified as growth promoting (428). However, these results are not consistent as LAR−/− mice have also exhibited reduced recovery following entorhinal cortex and sciatic nerve injury (419). Nonetheless, this complication means that the receptor may be growth promoting or inhibiting depending on a complex series of interactions. This, combined with the fact that LAR is not known to bind to CS-A or CS-C (267), may limit its use as a treatment for SCI and means more research is required into the mechanism of LAR action.

Recently, the Nogo receptors NgR1 and NgR3 (93) have also been recognized as CSPG receptors. Nogo−/− mice demonstrate only modest functional regeneration following injury (203, 353, 441), perhaps indicative that targeting this molecule alone is not sufficient to enable axon regeneration following SCI. Dickendesher et al. (93) recently demonstrated high-affinity binding between the receptors NgR1 and NgR3 and CSPGs, specifically CS-B, CS-D, CS-E, and HS (42, 267). Indeed, the binding patterns of RPTPσ, NgR1, and NgR3 fundamentally overlap. However, double knockout NgR1−/− and NgR3−/− mice showed significant axonal growth following injury not exhibited by either knockout alone (93). Regeneration was further enhanced in the triple NgR1−/−, NgR3−/−, and RPTPσ−/− knockout. This is highly suggestive of the important role that CSPG signaling has on regenerative capacity after traumatic injury.

D. Perineuronal Net Characterization

A key function of CSPGs within the adult CNS is their role in formation of PNNs (FIGURE 8, A and B). These extracellular reticular structures were originally identified by Camillo Golgi [reviewed by Celio et al. (64)]. The most widely addressed role of PNNs is their capacity to limit plasticity within the CNS. The PNN is activity dependent and forms around the soma and proximal neurites (FIGURE 8, B and C) appearing in the CNS at the time of synapse maturation and the closure of critical periods during development (148, 253). However, PNNs have additionally been implicated in a multitude of other homeostatic functions such as neuroprotection (43) and memory consolidation (141, 250, 308, 327). In addition, they are thought to play a role in the etiology of various neurological disorders such as addiction (354, 425), schizophrenia (254, 299, 300), Alzheimer’s disease (17, 305, 427), depression (150, 290), and epilepsy (298). While biochemical analysis suggests that only 1.3% of total CS-GAGs in the adult rodent brain are associated with PNNs (86), their critical role in the proper functioning of the CNS cannot be understated. While PNNs surround many neuronal subtypes, they are secreted in great abundance by parvalbumin-expressing GABAergic interneurons, many of which express the voltage-dependent potassium channel Kv3.1b (43, 164). They also surround fast-spiking interneurons (96, 274) and occur about many of the neurons in the spinal cord including motor neurons (253).

1. Perineuronal nets impede functional plasticity following SCI

The importance of PNNs to plasticity within the CNS is indicated by the rapid and robust degree of PNN remodeling that occurs in deafferented neuropil in centers distal to a lesion. For instance, Massey and co-workers (250, 251) showed that by 2–3 wk post injury, the PNNs were dramatically upregulated around rat dorsal column nuclei neurons following a mid-cervical dorsal column tract lesion that spared light touch sensory information from the thumb and forefinger (FIGURE 8B). Furthermore, evidence suggested that these upregulated nets were curtailing plasticity as their removal in the cuneate nucleus via ChABC allowed significant sprouting into its denervated portions from the intact sensory axons coming from the first two digits. Whether the spread of fibers throughout the nucleus improves or impedes function has not yet been determined. It has been demonstrated that thoracic SCI in a rodent model mediates decreases in parvalbumin expression and atrophy of GABAergic interneurons in hindlimb, sensory, and motor cortices (139, 140, 295). Furthermore, changes were noted in PNN CSPG-GAGs at the boundary between the motor forelimb and sensorimotor hindlimb cortex. Changes in the PNN may help tightly regulate the remodeling and plasticity that occurs in the cortex after SCI for the animal to acquire some degree of endogenous locomotor functional recovery.

PNNs also increase in areas surrounding motor neurons within the spinal cord well below the level of injury (295). An example of this occurs in the respiratory system. Alilain et al. (5) demonstrated that paralysis of the hemidiaphragm ipsilateral to a full C2 hemisection was persistent over time and leads to an increase of the PNN around phrenic motor neurons at level C4 (FIGURE 8C). However, acute removal of CS-GAGs at C4 by injection of ChABC enabled some, albeit minimal, return of function to the previously paralyzed side of the diaphragm muscle. We have recently assessed this treatment in animals up to 1.5 yr after injury where we hypothesized that a slow sprouting of re-crossed projections [the so-called latent crossed phrenic pathway; Alilain et al. (5)] over time, even in the presence of the PNN, might be unmasked by digesting the matrix via chondroitinase (410). We were surprised to see a rapid (as early as 1 wk) and robust recovery in diaphragm EMG activity in these long chronically paralyzed animals that was virtually indistinguishable from uninjured controls. Remarkably, the recovery after chronic stages was far greater than that which occurred at acute stages. In both studies, we show substantial reductions in both PNNs and CSPGs following ChABC application and sprouting of serotonergic axons, which are also critical in the control of breathing.

V. EXTRACELLULAR MATRIX-TARGETING REPAIR STRATEGIES

There are broadly three ways to remove the inhibitory influence of CSPGs (FIGURE 9). The first is to target the CSPG itself. The most commonly used method is via the enzymatic degradation of the CS-GAG chains which increases regeneration/plasticity at, or distant from, the site of injury. Other successful methods include preventing CSPG formation and inhibiting the proteoglycan signaling mechanism. These will each be described in turn.

FIGURE 9.

SCI repair strategies involving CSPG modification. A: CSPGs within the ECM and PNNs are upregulated following spinal cord injury. This molecule creates a physical and chemical barrier to the growth of axons. B: preventing the formation of CSPGs through processes such as xylosyltransferase inhibition, the use of xyloside, or N-acetylgalactosaminyltransferase-1 deletion removes the inhibitory component from the ECM enabling axonal growth. C: the catabolism of CS-GAGs through use of the bacterial enzyme ChABC removes the major inhibitory component of the glial scar and ECM facilitating axonal growth. The enzyme additionally facilitates plasticity through axonal and neuronal sprouting and synaptogenesis while encouraging neuronal protection. D: the inhibition of the CSPG receptor through the use of drugs such as ISP and ILP masks the inhibitory signals from the proteoglycans facilitating axonal and neuronal growth. E: combination treatment strategies may further facilitate repair following SCI, here demonstrating the use of ChABC with a peripheral nerve graft to facilitate axonal growth, sprouting, and plasticity as well as the formation of functional axonal connections.

A. Removing Chondroitin Sulfate-Glycosaminoglycans

The catabolism of CS-GAG chains to remove the inhibitory influence of CSPGs following SCI has the possible advantage of being a delayed strategy. As such, it can be applied at both acute and chronic time points after trauma, meaning that the positive effects of astrogliosis can occur before the treatment is applied. This would have the potential to maximize repair and recovery after SCI. Sulfatases, which target the catabolism of specific moieties of sulfated GAG chains such as CS-A, have been shown to enhance axonal regeneration in vitro (403) and in vivo (434). However, the most highly studied and successful of these methods is via the use of the GAG degrading enzyme ChABC. The bacterial enzyme acts to cleave GAG polymers from CS-A through E, into tetrasaccharides and disaccharides via an eliminative mechanism, preventing CSPG-matrix glycoprotein interactions (176, 313, 388). It was first applied in vitro and ex vivo to enable axonal outgrowth on CSPG substrates and glial scar explants (36, 256, 257, 329, 450). Indeed, ChABC treatment increases outgrowth over inhibitory astrocytic cell lines (356), through oligodendrocyte lineage cells (12), and facilitates neurite and axonal growth on spinal cord cryosections. The potential use of ChABC for in vivo SCI treatment was initially exemplified by Lemons et al. (230) who used the enzyme to degrade the CS-GAG chains of the glial scar following contusion. Subsequently, ChABC application was convincingly shown to mediate anatomical and functional regeneration with polysynaptic activity observed caudal to the injury site following a dorsal column crush (39).

The effects of ChABC upon axon regeneration following SCI have subsequently been demonstrated in a large number of studies using a variety of injury models and species including rats, hamsters, rabbits, cats, pigs, and squirrel monkeys. Indeed, ChABC enzyme-mediated axon regeneration has been shown in nigrostriatal (270), serotonergic (389), rubrospinal (433), and corticospinal tracts (39). Data suggest that ChABC can mediate recovery in the most severe of spinal injuries. Caggiano et al. (54) showed that intrathecal application of ChABC improved locomotor and autonomic bladder function following a clip compression injury. Furthermore, following contusion injury, ChABC has been shown to mediate some functional improvements in locomotor and urinary systems, albeit with modest anatomical regeneration (54, 193). However, the more dramatic effects of the use of ChABC have been in hemisected animals (180). Interestingly, Cafferty et al. (53) used a transgenic approach to express ChABC under a GFAP promotor. This enabled enzyme delivery to be targeted to areas of astrocytosis and demonstrated increased regeneration of CST axons into the site of a hemisection injury. Similarly, sensory and motor projections have shown functional growth into previously denervated areas (52, 368). These data may suggest that treatment with ChABC needs to occur at the optimal time and location after injury to evoke the maximal functional effect. Interestingly, recent studies have demonstrated that ChABC treatment leads to an increase of OPCs at the site of injury, meaning that treatment with the enzyme could facilitate remyelination following SCI (223, 307).

A critical effect of the application of ChABC is to promote plasticity from projections that survive after CNS trauma leading to functionally beneficial rewiring of undamaged systems in both rodent and non-human primate models (38, 134, 250, 251, 397). For example, numerous studies have shown that administration of the enzyme promoted serotonergic, corticospinal, and sensory fiber sprouting in areas where CS-GAG degradation was prominent (5, 23, 52, 193, 251, 402). Indeed, through anatomical tracing and electrophysiological techniques, collateral sprouting of forelimb sensory afferents into the partially denervated brain stem nuclei were shown to be functionally active following cervical dorsal column injury and ChABC injection (251). Similarly, Galtrey et al. (129) have shown that inaccurate repair following PNS injury can be overcome through ChABC injection into the spinal cord, causing increased neuronal sprouting. However, ChABC-induced sprouting does not always make functional connections (389). This highlights the need to drive the induced plasticity via appropriate activity and, thus, following Hebb’s postulate, stronger functional connections will be made (see below).

A number of studies have shown that ChABC can be neuroprotective and decrease CSPG-mediated axonal atrophy (62). Carter and co-workers (61, 62) demonstrated that CS-GAG removal by ChABC acts retrogradely to reduce degenerative changes of injured corticospinal and rubrospinal neurons, shown by a restoration in cell body soma size in acute and chronic injury models. This is an effect indicative of neuroprotection. Furthermore, ChABC has been shown to reduce macrophage-mediated axonal dieback following SCI (49). The precise mechanism through which ChABC mediates neuroprotective effects is not clear, although it may involve immune cell modulation (25, 94). However, these data clearly show that, if applied in the subacute stages following spinal trauma, ChABC activity can remove several of the negative effects of increased CSPG upregulation.

Despite ample evidence for the success of a ChABC-mediated treatment strategy for SCI, there are potential problems with its application. Primary among these are that the half-life of the enzyme is relatively short in vivo at ~3–10 days (342, 386) or 2–3 wk with glycerol, albumin, or trehalose stabilizers (67, 224, 280). This limits the time frame in which a single treatment application could evoke change in the CNS after injury. As such, either more invasive delivery procedures may need to be clinically employed, or the enzyme will need further modification. In a recent study, Lee et al. (224) produced a thermostabilized ChABC which remained active, albeit at low levels, up to 6 wk after SCI. However, this treatment has not demonstrated a functional recovery following SCI in vivo. Alternatively, a number of recent studies have utilized gene therapy to modify the enzyme to enable its secretion from mammalian cells (185, 206). Perhaps the most successful of these has shown that the modified enzyme can cause extensive in vivo CS-GAG removal resulting in CST sprouting following injury to the dorsal columns (276, 440). Furthermore, it has been shown that a lenti-viral form of ChABC can increase neuroprotection and enhance sensory-motor function from acute to chronic stages following spinal contusion injury (25). Despite more data being required, these studies demonstrate how innovations in ChABC can make it more applicable for clinical use.

While the CS-GAGs targeted through the use of ChABC remove the major inhibitors, the core protein and the leftover ChABC-resistant sugar stubs remain (337). Another way to remove CSPGs, and specifically target the core protein following SCI, is to take advantage of the endogeonous CNS CSPG regulators. There are a number of identified proteases that are able to mediate this function and act to degrade the proteoglycan core proteins. Perhaps the most widely studied is the one known to act all over the body, the disintegrin-like and metalloproteinase with thrombospondin type 1 motif 4 (ADAMTS4) (9, 227, 383). It is thought that experience-dependent plasticity within the CNS is mediated through a tPA-ADAMTS4 signaling pathway (252, 294). Functional regeneration may be increased following SCI through the upregulation of ADAMTS4 expression by the introduction of IL-1α (228). Alternatively, microarray studies have identified a host of MMPs, which act in the CNS to breakdown CSPGs (190). These molecules are known to be upregulated in response to neuronal stimulation (163, 248) and can aid recovery following SCI (222, 228).

The problem with increasing MMPs or ADAMTS4 activity and expression after SCI is the method through which they operate. Following SCI, CSPG production changes the ECM which expresses a higher quantity of CS-GAGs. These chains provide a significant physical barrier to the action of both MMPs or ADAMTS4 meaning that their functional activity following SCI may be limited. However, recent studies have shown that combining ADAMTS4 activity with ChABC application can have an additive effect on neurite outgrowth in vitro (78). These data illustrate how full degradation of the entire CSPG in a spatially and temporally specific manor may mediate neuronal growth following injury and highlight this combination as a potential SCI treatment strategy.

B. Preventing Chondroitin Sulfate-Glycosaminoglycan Formation

An alternative method to prevent the inhibitory effect of CSPGs following SCI is not to degrade them once in place, but to prevent their formation from the beginning. However, this has the disadvantage of needing to be applied within the acute phase of injury, limiting clinical applicability. The classic method of preventing CSPG formation involves the use of a DNA enzyme to inhibit xylosyltransferase (XT)-1 mRNA. This enzyme glycosylates the core protein of CSPGs, and without its functioning, proteoglycan deposition is predominantly inhibited and axon regeneration can be stimulated (144). However, some CSPG synthesis was evident in this experiment about the lesion penumbra, largely attributed to the function of XT-2.

To have the benefit of reduced CSPG allowed plasticity and axonal growth after trauma, many groups have focused on inhibiting the formation of CS-GAGs while enabling the core protein to be expressed. For example, Takeuchi et al. (376) have recently described the use of a transgenic animal which prevented CS synthesis through deletion of N-acetylgalactosaminyltransferase-1. These mice showed superior recovery from SCI than both control and ChABC-treated animals, possibly due to the endogenous upregulation of HS. Similarly, the use of the molecule xyloside to compete with the attachment of the CS-GAGs to the CSPG protein core reduces the inhibitory effects of the proteoglycan on neuronal growth by decreasing their deposition about the site of injury in SCI and multiple sclerosis models (223, 449). A number of molecules that have a similar function on CS formation have likewise been used with comparable effects including 4-fluoro-glucosamine (282) and 4-methyl-umbelliferyl-β-d-xylopyranoside (340, 341). Similarly, prevention of GAG polymerization using siRNA against chondroitin polymerizing factor has been shown to successfully decrease the CS-GAGs expressed and secreted by astrocytes, facilitating the growth of cerebellar granule cell neurites (218). Keough et al. (198) have recently described the generation of a new CSPG synthesis inhibitor, fluorosamine. This molecule likely interferes with the enzymatic conversion of UDP-N-acetylglucosamine to UDP-N-acetyl-galactosamine by 4-epimerase (282). CSPGs have long been known to inhibit OPC differentiation, process outgrowth, and remyelination (198, 223, 348). Fluorosamine acted in vivo to reduce CSPGs, in turn facilitating OPC process outgrowth and remyelination in models of multiple sclerosis.

C. Manipulating Chondroitin Sulfate-Glycosaminoglycan Receptors and Signaling

The final method employed to evoke recovery from SCI by negating the inhibitory influence of CSPGs is to target the proteoglycan signaling mechanism. Again, this treatment strategy has the advantage of being able to be applied at both acute and chronic stages after injury. Lang et al. (221) demonstrated that RPTPσ activation in adult sensory neurons by a gradient of CSPGs directly caused growth cones to enter an entrapped dystrophic state. Furthermore, it was shown that the CSPG receptors RPTPσ and LAR occurred at high concentrations within these dystrophic growth cones. The researchers designed an innovative peptide-mimetic for the PTPσ wedge termed the intracellular sigma peptide (ISP) which enabled a change in the growth cone even after it had entered an endball state, allowing for advancement across the CSPG-laden territory. Use of the peptide delivered systemically following severe thoracic contusion resulted in remarkable functional recovery of the locomotor and urinary systems, possibly mediated, in part, through serotonergic sprouting and reinnervation caudal to the site of injury (221). Recently, Paveliev et al. (304) have used the CS-binding protein pleiotrophin (a heparin-binding growth-associated molecule) to facilitate neural regeneration in vitro and in vivo. In preventing CS binding to PTPσ, they demonstrated increased sprouting following mild cortical and spinal cord injury, although this may additionally be linked to rapid reestablishment of the blood supply following trauma. Inhibition of the LAR receptor through the peptide ILP (114, 420) has been shown in vitro to promote axonal growth on CSPG substrates by activating Src, FAK, and TrkB as well as a number of downstream molecules (114, 221, 428). In vivo, cord injured mice were shown to have superior locomotor function than controls, again possibly caused by increased axonal sprouting of serotonergic fibers caudal to the site of injury (114). These data hold great promise for the treatment of SCI and, potentially, a number of neurological conditions in which CSPGs play a definitive role. Furthermore, these exciting results provide great incentive for the identification of receptors for CS-A and CS-C which, if inhibited, may provide further functional benefit after CNS trauma.

D. Combinatorial Strategies

Because SCI is a multifaceted problem, it is unlikely that any one treatment alone can affect the large number of obstacles that occur following trauma. Indeed, this may be the reason why many disparate treatments produce similar modest levels of recovery in SCI animal models. Thus combining treatment strategies that target contrasting mechanisms is being investigated by many laboratories in an attempt to maximize functional benefit. To this end, ChABC has been used widely in combination treatment strategies to aid functional recovery. For example, Bai et al. (16) utilized ChABC with the β2-adrenoceptor agonist clenbutarol following complete T10 transection to increase cAMP levels, facilitating neuroprotection. Only the combined treatment group showed enhanced anatomical and functional recovery 2–3 mo following injury. The use of ChABC to remove the inhibitory CSPG barrier preventing neuronal growth has been used most effectively in combination with tissue grafts and peripheral nerve bridges. The enzymatic modulation of the glial scar at the entrances to the bridge, in particular with the further addition of FGF, facilitates integration of the transplanted tissue with the host and facilitates the regrowth of axons through and beyond the graft tissue (88). Tom et al. (391) demonstrated this through the application of ChABC with a peripheral nerve graft (PNG) after chronic contusion injury, demonstrating the formation of functional synapses and some behavioral recovery. This strategy has been further developed through the addition of virally mediated exogenous BDNF to the combination (390). While the number of regenerating axons did not increase, the combination did enhance the number of neurons trans-synaptically activated following stimulation of the graft. Similarly, the combination of ChABC with a PNG and acidic fibroblast growth factor has been used in rodent models following complete T8 transection to mediate remarkable axonal regeneration from brain stem centers leading to recovery of urination (88). Furthermore, the use of ChABC combined with a PNG has been used to facilitate functional recovery of the respiratory motor system following acute cervical hemisection (5). Three months following the injury, the combined treatment group showed the maximal return of diaphragm function and experiments suggested that the treatment strategy further led to a beneficial rewiring of respiratory motor circuitry. The positive effects of combining ChABC treatment with grafting techniques have been shown in numerous other publications and animal models (118, 119, 173, 193, 226, 400).

In addition to causing the catabolism of the glial scar, ChABC can act to increase plasticity within the injured spinal cord below the level of injury especially when combined with targeted exercise strategies. It has been shown that combining ChABC with task-specific physical rehabilitation following acute and subchronic injury can augment this plasticity, mediating specific functional recovery (134, 402). Combining ChABC treatment with other strategies which reduce inhibition in the extrinsic environment of the spinal cord has also yielded some success. Zhao et al. (440) showed that the combination of ChABC, rehabilitation, and anti-Nogo-A treatment increased the degree of axon sprouting and regeneration better than either treatment acting alone. Furthermore, Wang et al. (406) demonstrated that ChABC added to preconditioning with a sciatic nerve lesion and use of a soluble Nogo receptor decoy (NgR310-Fc) enabled axonal growth to extend not only through, but beyond the injury site. It will be exciting to learn whether the various systemically delivered treatment strategies for overcoming CSPG-mediated inhibition when combined with appropriate rehabilitation will produce enhanced recovery sufficiently robust to provide hope for clinical translation especially at chronic stages after injury.

VI. CONCLUDING REMARKS

The etiology of regeneration failure following SCI has evolved from a simplistic view that the astrocyte scar is the solely culpable mechanical obstacle to axonal regrowth to a richer, multicellular and molecular perspective that involves astrocytes, immune cells, perictyes and fibroblasts as well as oligodendrocyte progenitors, and the intricate ways they all interact to form an axon entrapping environment. While this review concentrated on the potently inhibitory effects of upregulated CSPGs in the scar as well as the PNN and the emerging ways this family of proteoglycans could be perturbed to promote regeneration or sprouting, understanding the complexity of the glial/fibrotic scar and net has underscored the necessity for combinatorial treatments. In particular, the fibrotic component of scar, produced by pericyte and/or meningeal fibroblast associations with astrocytes and each other, and manipulations to specifically alter these interactions have been relatively less well studied than the glial cells themselves. From our perspective, we would like to see the field further pursue combinatorial treatment paradigms especially targeted to the chronic stage since the millions of people worldwide who suffer from SCI are well past their moment of injury. The rather daunting problems that the long-injured cord introduces to hinder the regrowth of damaged axon pathways has frustrated attempts to devise promising regenerative approaches. The exciting discovery of the first known receptors on neurons that mediate the regrowth inhibitory effects of the entrapping matrix molecules has opened the door to the production of specific blocking peptides as well as long-acting enzymes that can be used to manipulate these extracellular culprits and their receptors. When combined with additional strategies to enhance intrinsic neuronal growth potential (238) as well as targeted rehabilitative therapy (158, 264), we may be able to elicit recovery even after a near lifetime of paralysis.

GRANTS

P. M. Warren was funded by the Welcome Trust ISSF Fellowship with the University of Leeds. P. M. Warren and J. Silver were funded by the International Spinal Research Trust, Wings for Life, and the Neilsen Foundation. A. P. Tran and J. Silver were funded by National Institute of Neurological Disorders and Stroke Grant NS025713, The Hong Kong Spinal Cord Injury Fund, The Brumagin/Nelsen Fund, and The Kaneko Family Fund.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.