Chondroitin sulphate proteoglycans: preventing plasticity or protecting the CNS?

Abstract

It is well established that axonal regeneration in the adult CNS is largely unsuccessful. Numerous axon-inhibitory molecules are now known to be present in the injured CNS, and various strategies for overcoming these obstacles and enhancing CNS regeneration have been experimentally developed. Recently, the use of chondroitinase-ABC to treat models of CNS injury in vivo has proven to be highly beneficial towards regenerating axons, by degrading the axon-inhibitory chondroitin sulphate glycosaminsoglycan chains found on many proteoglycans in the astroglial scar. This enzyme has now been shown to restore synaptic plasticity in the visual cortex of adult rats by disrupting perineuronal nets, which contain high levels of chondroitin sulphate proteoglycans (CS-PGs) and are expressed postnatally around groups of certain neurons in the normal CNS. The findings suggest exciting prospects for enhancing growth and plasticity in the adult CNS; however, some protective roles of CS-PGs in the CNS have also been demonstrated. Clearly many questions concerning the mechanisms regulating expression of extracellular matrix molecules in CNS pathology remain to be answered.

Introduction

Axonal regeneration is poor in the adult CNS

Injured neurons of the adult mammalian central nervous system (CNS) have a limited capacity to regenerate their axons, which rarely succeed in reaching the target and in forming new functional synapses. Consequently, damage to the CNS is normally irreparable. By contrast, axons of the injured peripheral nervous system (PNS) have a strong capacity for regeneration and functional recovery.

What is the reason for this dichotomy? Successful axonal regeneration depends on several factors: survival of the cell body; axonal sprouting from the proximal stump of the lesioned neuron; elongation of these axonal sprouts towards their original target; axonal ensheathment; removal of degenerating axons and eventually restoration of contacts between axon terminals and their target. Whilst these events occur successfully in the PNS (see Bray et al. 1981 for a review), with structural and functional recovery of the nerve, lesioned axons of the CNS generally exhibit only limited sprouting; these sprouts do not elongate and regeneration fails.

A combination of several causes is believed to underlie this failure. CNS axons have an intrinsically reduced capacity for growth: embryonic or early postnatal axons are considerably more able to regenerate following a lesion than those of adults (Kalil & Reh, 1982; Schreyer & Jones, 1983). The CNS environment is clearly less conducive to axonal growth than that of the PNS; the glia of the two environments show a number of differences. The PNS is rich in Schwann cells, and axons from both PNS and CNS neurons grow effectively in such an environment (Richardson et al. 1980, 1984; David & Aguayo, 1981; for review see Aguayo et al. 1981). The CNS, however, contains astrocytes and oligodendrocytes through which both PNS and CNS axons grow poorly (Fawcett et al. 1989a, b; Bandtlow et al. 1990), implying glia and/or other elements of the adult CNS environment are not suitable for axonal regeneration – owing to expression of inhibitory molecules and/or insufficient expression of guidance and trophic factors. Additionally, after damage to the CNS tissue, axonal and glial debris remain for many months and could be a source of axon inhibition. By contrast, debris is cleared from lesions much more rapidly in the PNS (Perry et al. 1987; Stoll et al. 1989).

A number of studies have been carried out over the last two decades to isolate molecular differences between the CNS and PNS environments; numerous potentially axon-inhibitory molecules are found in the CNS (discussed below). Furthermore, where the adult CNS is injured a scar is formed: it is in this scar tissue that axon regeneration fails (Davies et al. 1997). Many of the axon growth-inhibitory molecules have been shown to be up-regulated or only present in the astroglial scar of the lesioned CNS. This astroglial scar is specific to the mature CNS (Berry et al. 1983), and represents a major impediment to growing axons. Although the scar takes days or weeks to mature, it begins to form very rapidly after injury, resulting in a densely packed, highly complex structure consisting of reactive glial cells and up-regulated extracellular matrix (ECM; see below), which surrounds the lesion site.

Growth-inhibitory molecules in the CNS

A variety of molecules known to exist in the adult CNS have been shown through in vitro and in vivo studies to demonstrate axonal growth-inhibitory properties. More importantly, many of these molecules are specific to, or up-regulated in, the injured CNS environment. The molecules may be loosely classified into two main types: myelin-associated molecules and ECM constituents.

Myelin-associated molecules

In the late 1980s Schwab and colleagues demonstrated that a substrate non-permissive to growth exists in CNS white matter (Schwab & Thoenen, 1985; Schwab & Caroni, 1988; Crutcher, 1989; Savio & Schwab, 1989), leading to the hypothesis that where myelinated axons are disrupted, debris containing myelin-associated axon-inhibitory molecules will be present around the lesion. A number of myelin-associated molecules have since been isolated that have axon-inhibitory properties, including the two neurite growth inhibitors Nogo (originally called NI-250) and myelin-associated glycoprotein (MAG). Much literature has been published recently concerning MAG, Nogo and the Nogo receptor (Nogo-66) that is beyond the scope of this article; for recent reviews see Bandtlow (2003) and McGee & Strittmatter (2003).

ECM constituents

For many neurons, their migration and axon elongation occurs through the ECM: in the CNS the ECM is largely devoid of cells but contains several types of molecules with which neurons and glia interact, and these can have important influences on many aspects of a cell's behaviour.

The extracellular space surrounding many non-neuronal cells in the CNS is filled with a network of glycoproteins, proteoglycans and hyaluronan; close to the membrane of such cells the ECM becomes more dense, forming a basement membrane composed principally of collagens; glycoproteins – particularly tenascin-C and tenascin-R; chondroitin sulphate proteoglycans (CS-PGs) and heparan sulphate proteoglycans (HS-PGs); hyaluronic acid (HA); cell adhesion molecules and integrins.

ECM molecules expressed in the developing CNS may have both growth-promoting and growth-inhibitory effects on axons (e.g. Snow et al. 1990b; Oakley & Tosney, 1991; Brittis et al. 1992; Emerling & Lander, 1996; Treloar et al. 1996; for review see Margolis & Margolis, 1997). Many of these molecules are up-regulated in the adult ECM following a lesion to the CNS (in particular CS-PGs; see below), and have been shown to have inhibitory properties towards regenerating axons.

Tenascin is an important secreted ECM component with a range of binding sites and functions (Hoffman et al. 1988; Steindler et al. 1989; Grumet et al. 1994; Husmann et al. 1995; for review see Faissner, 1997). Tenascin is abundant in the basement membrane, being produced by astrocytes during development, with important roles in mediating axon–glia interactions (Steindler et al. 1989; Faissner & Kruse, 1990; Lochter et al. 1991). There are two members of the tenascin gene family: tenascin-C and tenascin-R; tenascin-C is expressed as numerous alternatively spliced variants with various functions (Faissner et al. 1988; Stern et al. 1989; Chuong & Chen, 1991; Faissner, 1997). The same tenascin molecule may have either growth-inhibitory or growth-promoting effects towards different neurons within different contexts; a number of studies have demonstrated the neurite growth-inhibitory properties of tenascin in vitro (Pesheva et al. 1989; Crossin et al. 1990; Faissner & Kruse, 1990). It also has growth-promoting effects ascribed to the alternatively spliced A-D and D5 domains (Meiners et al. 1999, 2001).

Tenascin is found in the normal adult CNS although at lower levels than in the developing CNS. Production is up-regulated in the glial scar after injury (Laywell et al. 1992); the increased tenascin has been co-localized with reactive glial fibrillary acidic protein (GFAP)⁺ astrocytes (Lochter et al. 1991; McKeon et al. 1991; Laywell et al. 1992). Tenascin is known to interact with many CS-PGs in vitro (Grumet et al. 1994; Xiao et al. 1997, 1998), and thus may be capable of forming (axon-inhibitory) complexes with CS-PGs in injured CNS tissue.

The astroglial scar

A lesion to the CNS provokes reactive behaviour among local glia, which proliferate and/or migrate to the lesion site, forming the astroglial scar (Fig. 1). These cells – principally astrocytes, microglia and oligodendrocyte precursor (OP) cells – respond rapidly: they may be identified by their hypertrophied cell body and thickened processes, and altered production of a number of cell surface receptors and ECM molecules, including tenascin and proteoglycans (McKeon et al. 1991, 1995; Levine, 1994; Haas et al. 1999; Asher et al. 2000; see reviews by Streit & Kincaid-Colton, 1995; Ridet et al. 1997; Raivich et al. 1999). Reactive microglia and OP cells also undergo local proliferation: these cells are recruited around the lesion site in large numbers (Levine, 1994; Rhodes et al. 2003), where they contribute to the astroglial scar. These changes begin within hours and develop over about 7–10 days, leading eventually to the classic appearance of the mature glial scar, which consists of highly branched astrocytic processes attached to one another by junctional processes, with plentiful ECM (Maxwell et al. 1990), surrounding macrophages, fibroblasts and reactive glia at the lesion core.

Fig. 1.

Simplified diagram showing major functions, markers expressed and molecules produced by glial cells in (A) the normal adult CNS and (B) the lesioned CNS (direct trauma). Morphological, phenotypical and functional changes are observed among glial cells contributing to glial scar formation around the lesion site; oligodendrocytes are not included. Inflammatory cells (monocytes, neutrophils and T cells) would also be recruited across the compromised blood–brain barrier (not illustrated), producing a range of pro-inflammatory chemotactic cytokines and chemokines, assisting endogenous microglia with antigen recognition and phagocytotic functions, and stimulating further gliosis. Bold type indicates new expression of a protein; ↑ = up-regulated expression of a protein; scale bars = 5 µm. Images taken from adult rat basal ganglia in normal brain (A) and 24 h following microinjection of 5 µg lipopolysaccharide (B) causing rapid inflammatory response and gliosis in microglia (labelled with antibodies to CR3), astrocytes (labelled with antibodies to GFAP) and OP cells (labelled with antibodies to NG2). Abbreviations not in text: CAM, cell adhesion molecule; CR3, complement receptor 3; FGF, fibroblast growth factor; MCSF(-R), macrophage colony stimulating factor (receptor); MHC, major histocompatibility complex; NGF, nerve growth factor; PDGFαR, platelet-derived growth factor receptor-alpha; TGF, transforming growth factor; TNF, tumour necrosis factor. Based on information from Ridet et al. (1997), Raivich et al. (1999) and Asher et al. (2001).

Chondroitin sulphate proteoglycans

Significant studies by Davies et al. (1997, 1999) demonstrated that adult dorsal root ganglion (DRG) neurons microtransplanted into undamaged adult white matter tracts were able to grow through this CNS environment. However, these axons stopped growing upon reaching damaged CNS tissue. Immunohistochemistry revealed that the ECM-rich gliotic scar surrounding the damaged tissue contained a large proportion of CS-PGs, the distribution of these molecules matching the regions of cessation of axon growth. These and other studies suggest that glial scar tissue is more inhibitory to axon regeneration than undamaged adult CNS tissue, and show a correlation between inhibition of axon growth and the presence of elevated levels of CS-PGs.

Proteoglycans: general structure

Proteoglycans encompass a wide range of complex molecules found throughout different tissue types; they have been associated with a variety of cellular processes including cell adhesion, growth, receptor binding, cell migration, barrier formation and interaction with other ECM constituents (for review see Bandtlow & Zimmerman, 2000). These diverse molecules comprise a core glycoprotein with glycosaminoglycan (GAG) sugar chains covalently attached. Each GAG consists of a simple, linear polymer of repeating disaccharide units, composed from two alternating monosaccharides: usually one sugar is a uronic acid and the other is either N-acetylglucosamine or N-acetylgalactosamine. Different types of GAG may be created as a result of sulphation and epimerization modifications that are carried out on the sugars themselves following polymerization. The length of GAG chains may also vary, from a polypeptide chain of just 10 kDa to 400 000 kDa, and the core protein itself may have varying numbers of GAG chains attached – from one to well over 100, resulting in a diverse array of PGs with a range of functional complexities. Many of the functional properties of PGs are attributed to the attached side chains; much of the interaction between PGs and cell-surface receptors or ECM proteins is thought to occur via binding sites on the GAG chains, although the core protein is also able to bind substrates (see Bandtlow & Zimmerman, 2000).

CS-PG expression patterns and functions in the CNS

The CNS is rich in proteoglycans, particularly in those containing sulphated GAG, such as CS-, HS-, dermatan sulphate (DS)- and keratan sulphate (KS)-PGs. CS-PGs are characterized by sulphation of the GAG side chains at certain positions on the sugars; these GAG chains may be cleaved from the core protein by chondroitinase-ABC (or chondroitinase-AC).

The CS-PGs are represented by several types, including large aggregating proteoglycans such as aggrecan (Paulsson et al. 1987) and versican (Krusius et al. 1987); the brain-specific proteoglycans neurocan (Grumet et al. 1993, 1994; Friedlander et al. 1994; Oohira et al. 1994) and brevican (Yamada et al. 1994); and NG2 (Levine & Card, 1987; Stallcup & Beasley, 1987) and phosphacan/DSD-1 (Grumet et al. 1993, 1994; Maeda et al. 1994; Maurel et al. 1994). These CS-PGs are abundant within the ECM; all are expressed in the CNS, and have been shown to interact with a variety of other matrix components including laminin, fibronectin (Schmidt et al. 1991), tenascin (Grumet et al. 1994), HA (Krusius et al. 1987; Paulsson et al. 1987; Doege et al. 1991) and collagen (Bidanset et al. 1992; Hedbom & Heinegard, 1993). Interactions of CS-PGs with cell surface receptors or other ECM components may occur through the GAG chains or via the core protein.

CS-PG expression is particularly strong in the embryonic brain: the patterns of expression during development (in regions such as neural crest pathways and spinal cord) (Snow et al. 1990a; Oakley & Tosney, 1991) and axon-inhibitory properties are consistent with the theory that CS-PGs behave as axon guidance molecules during development, because growing axons generally avoid regions rich in CS-PGs (Snow et al. 1990b; Oakley & Tosney, 1991; for review see Asher et al. 2001), although there are some reports of axon-promoting mechanisms of CS-PGs (Brittis et al. 1992; Emerling & Lander, 1996; for review see Margolis & Margolis, 1997). Furthermore, it has been demonstrated in vitro that removal of chondroitin sulphate from models of developmental axon guidance abolishes the inhibition (e.g. Brittis et al. 1992).

The expression patterns of CS-PGs are altered in the adult CNS. For example, neurocan is mainly present in the white matter in the adult CNS; this protein is produced largely by astrocytes but can also be produced by OP cells (Asher et al. 2000). NG2 is distributed throughout the adult brain on OP cells (Levine & Card, 1987; Levine et al. 1993), on some blood vessels and meningeal cells (Morgenstern et al. 2002); this molecule may additionally be cleaved and secreted into the ECM (Nishiyama et al. 1995). Phosphacan/DSD-1 and the membrane-associated form, RPTPβ, is present at particularly high levels in Bergmann glia and nuclei in the cerebellum of the postnatal rat, and elsewhere in the brain and spinal cord (Meyer-Puttlitz et al. 1996). Versican is found in the white matter, and produced by OP cells (Asher et al. 2002); brevican may be found distributed throughout the CNS (Seidenbecher et al. 1995), produced by astrocytes (Yamada et al. 1997). Versican, brevican and aggrecan are all thought to be expressed in greater quantities in the normal adult brain than in the developing brain (Milev et al. 1998).

CS-PGs inhibit neurite growth in vitro and in vivo

Many studies have demonstrated in vitro that CS-PGs are inhibitory towards neurite outgrowth, either via their CS chains or their core proteins (Dou & Levine, 1994, 1997; Friedlander et al. 1994; Smith-Thomas et al. 1994; Yamada et al. 1997; Niederöst et al. 1999; Asher et al. 2000; Schmalfeldt et al. 2000). Additionally, both phosphacan and neurocan are able to bind with high affinity to the cell adhesion molecules N-CAM and Ng-CAM/L1 (Grumet et al. 1993; Friedlander et al. 1994), thus interfering with their interactions and indirectly inhibiting neurite outgrowth (Friedlander et al. 1994; Milev et al. 1994).

Reactive astrocytes and/or OP cells within the glial scar (see Fig. 1) have been shown to up-regulate their expression of various axon-inhibitory CS-PGs in vivo, in particular neurocan (Haas et al. 1999; McKeon et al. 1999; Asher et al. 2000), NG2 (Levine, 1994; Ong & Levine, 1999; Rhodes et al. 2003), phosphacan (Laywell & Steindler, 1991; McKeon et al. 1995) and versican (Asher et al. 2002; Jones et al. 2003; Tang et al. 2003). Several studies have identified an up-regulation of inhibitory CS in CNS lesion sites, using the antibody CS56, which binds to CS chains on proteoglycans (e.g. McKeon et al. 1991; Davies et al. 1997, 1999; Fitch & Silver, 1997).

Overcoming CS-PG inhibition of axon growth: chondroitinase-ABC

It has been established in vitro that much of the axon growth-inhibitory properties of several CS-PGs reside in the CS side chains because their inhibitory effects may be abolished with the enzyme chondroitinase-ABC, or by interfering with CS GAG synthesis (Friedlander et al. 1994; Milev et al. 1994; Smith-Thomas et al. 1995; Schmalfeldt et al. 2000). NG2 is an unusual CS-PG in that the inhibitory effects of the purified molecule on neurite outgrowth are not affected by chondroitinase (Dou & Levine, 1994); however, the use of NG2 neutralizing antibodies has been shown to abolish inhibitory effects of NG2-containing cell membranes (Chen ZJ et al. 2002).

Using an ex vivo model of the glial scar (Rudge et al. 1989), adult scar tissue was shown to be rich in reactive astrocytes and CS-PGs, which were inhibitory towards neurite outgrowth in vitro; treatment with chondroitinase was able to overcome this inhibition (McKeon et al. 1991, 1995).

More recently chondroitinase has been applied to models of CNS injury in vivo. Treatment of the lesioned adult rat nigrostriatal system (Moon et al. 2001) or dorsal column (Bradbury et al. 2002) with chondroitinase degraded CS-GAG around the lesion site and promoted axonal regeneration. Bradbury et al. (2002) demonstrated an accompanying increase in the expression of the growth-associated protein (GAP)-43 in lesioned neurons, restoration of post-synaptic activity below the lesion and functional recovery. The beneficial effects of chondroitinase towards aiding axonal regeneration after injury have been shown in further recent studies (Krekoski et al. 2001; Zuo et al. 2002; Tropea et al. 2003; Yick et al. 2003).

Protective roles of CS-PGs in the CNS

CS-PGs are up-regulated where the blood–brain barrier breaks down

It is currently not known precisely what triggers gliosis and eventual formation of the astroglial scar, the end product of a cascade of events. It has been observed, however, that up-regulation of CS-PGs may be immunohistochemically localized to the region of tissue damage and inflammation, and hence compromise of the blood–brain barrier (BBB) (Fitch & Silver, 1997; Fitch et al. 1999; Rhodes et al. 2002; our unpublished personal observations).

Some studies have suggested candidate molecules involved in promoting gliosis as a result of BBB breakdown: in particular, the serine protease thrombin has been implicated. Thrombin, derived from its zymogen prothrombin and an important part of the blood coagulation cascade (see Davie et al. 1991, for review), has multiple functions including activation of platelets, and chemotaxis and adhesion in monocytes, macrophages and neutrophils (reviewed by Grand et al. 1996). Thrombin may also influence cells in the CNS, including regulating neurite outgrowth (Gurwitz & Cunningham, 1988), inducing astrocytic proliferation (Grabham & Cunningham, 1995) and microglial activation (Möller et al. 2000). Thrombin and other serine proteases would extravasate across a compromised BBB; however, the rapid effects of thrombin on CNS cells demonstrated in vitro are not mimicked by other serine proteases (see Grand et al. 1996). Nishino et al. (1993) infused thrombin into rat brains for 7 days and provoked significant inflammation and gliotic reaction, including infiltration of inflammatory cells; proliferation; reactivity of astrocytes and induction of angiogenesis. In support of these findings, mRNA to both prothrombin and one of the thrombin receptors (belonging to the family of protease activated receptors (PAR) 1–4), PAR-1, is found widely distributed throughout the brain in neurons and glia (Dihanich et al. 1991; Weinstein et al. 1995; Niclou et al. 1998; see review by Gingrich & Traynelis, 2000). Other candidate molecules suggested to trigger gliosis include pro-inflammatory cytokines such as interleukin (IL)-1 (Giulian & Lachman, 1985; Giulian et al. 1988) and ciliary-derived neurotrophic factor (CNTF) (Kahn et al. 1995; Winter et al. 1995). Although all of these molecules clearly play important roles in orchestrating the CNS injury response, none has yet been established as the primary trigger of gliosis; it is likely that a combination of factors is required in order to initiate up-regulation of CS-PGs at sites of tissue damage.

Whilst astrocytes and microglia may become reactive following lesions affecting the CNS where the BBB remains intact (for example around centrally located cell bodies of peripherally lesioned axons, or at the distal or proximal ends of injured neurons; see Raivich et al. 1999), it would appear that such reactive astrocytes do not alter their expression patterns of CS-PGs, although their phenotypes appear otherwise reactive (see Ridet et al. 1997). In keeping with this, Fitch & Silver (1997) demonstrated that increased CS-PG immunoreactivity correlated with evidence of BBB breakdown but not necessarily with astrogliosis further from the lesion. Additionally, observations from within our laboratory suggest that OP cells only become reactive where the BBB is compromised, these cells rapidly increase their expression of NG2, and take on a reactive morphology (Rhodes et al. 2002).

Sobel & Ahmed (2001) correlated increased expression of the CS-PGs aggrecan, neurocan and versican with increased astrocytosis in the pro-inflammatory environment surrounding multiple sclerosis (MS) plaques in human patients; these alterations to the ECM were believed to be a result of the compromised BBB, and ensuing release and activation of proteases. Interestingly, when Fitch and colleagues examined injured rat CNS tissue, it was found that cavities which had developed near to the original wound, as made evident by strong inflammatory reactions, were equally demarcated by up-regulated CS-PG expression (Fitch & Silver, 1997; Fitch et al. 1999).

These findings suggest proteoglycans surrounding regions of damaged tissue and disrupted BBB have a vital role in sealing off the lesioned area and perhaps limiting further cavitation and secondary injury. Such a mechanism would rely on the very rapid ability for local glial cells to increase their production of CS-PGs in response to factors either present in extravasing serum, or on infiltrating inflammatory cells.

Although further understanding of the triggers of increased CS-PG synthesis at sites of CNS tissue damage might enable development of tools that would allow greater manipulation of the axon-inhibitory glial scar, the findings discussed above suggest that preventing up-regulation of CS-PGs at the site of a wound might have detrimental effects on the healing and sealing processes.

Proteoglycans and models of disease

Given the abundancy of proteoglycans in ECM and connective tissues, it is perhaps unsurprising that these molecules have been implicated in a diverse range of syndromes or diseases. Investigations of pathologies of connective tissues (including cartilage and skeletal systems, particularly forms of arthritis) have associated the pathology with degradation of the ECM and a resulting lack of DS-/CS-PGs that are an important component of the articular cartilage (e.g. Richardson, 1985; Pulkkinen et al. 1990; Blackshear et al. 1997; for review see Aigner & McKenna, 2002). More recently, models of experimentally induced proteoglycan deficiencies have linked mutant mice lacking various proteoglycans with abnormal development and pathological phenotypes (e.g. Xu et al. 1998; Serpinskaya et al. 1999; Baeg et al. 2001; Ameye et al. 2002; Chen XD et al. 2002; Jepsen et al. 2002) (for recent reviews see Ameye & Young, 2002; Schwartz & Domowicz, 2002).

There has also been some evidence of neuroprotective roles of CS-PGs within the CNS. For example, recent studies examining the mechanisms of beta amyloid-induced neurodegeneration in models of Alzheimer's disease have noted a correlation between neurons that are normally devoid of perineuronal nets (PNNs; see below) and their susceptibility to Alzheimer's disease (Brückner et al. 1999; Adams et al. 2001; Hartig et al. 2001), suggesting that the high concentration of CS-PGs surrounding neurons with PNNs can somehow protect them from the formation of Alzheimer's lesions. The initiation of neurodegenerative processes has been associated with abnormal deposition of HS-, DS- and CS-PGs and increased proteolytic cleavage of ECM components (Belichenko et al. 1999). Studies in vitro have additionally shown neuroprotective effects of chondroitin sulphate: CS-PGs were added to cultures of rat hippocampal or cortical neurons in order to rescue the cells from excitotoxic damage (Okamoto et al. 1994a, b); Woods et al. (1995) attenuated beta-amyloid-induced neurodegeneration of hippocampal cultures using CS- and HS-GAG. Chondroitin sulphates are also able to protect chondrocytes from inflammation and articular symptoms in models of osteoarthritis both in vitro (Nerucci et al. 2000) and in vivo (Uebelhart et al. 1998); this has been reviewed by Bali et al. (2001).

Pathological implications for HS-PGs

Interestingly, the HS-PGs seem to have been more widely implicated with human pathology than CS-PGs. Whilst the proteoglycan form of amyloid precursor protein (APP), appican, does contain a CS chain (Shioi et al. 1992, 1993) and CS-PGs have been implicated in Alzheimer's disease (e.g. De Witt et al. 1994, 1994), the majority of papers demonstrate increased expression of HS-PGs in neurodegenerative plaques (e.g. Schubert et al. 1988; Snow et al. 1988; Su et al. 1992; Caceres & Brandan, 1997; for review see Bornemann & Staufenbiel, 2000). HS-PGs have also been shown to promote phosphorylation of the microtubule associated protein tau (Goedert et al. 1996); Canning et al. (1993) showed beta amyloid in the brain stimulates astrocytes to become reactive and increase their deposition of CS- or HS-PGs.

Much literature has also associated up-regulation of HS-PGs (syndecans and glypicans in particular) with invasive gliomas (e.g. Steck et al. 1989; Liang et al. 1997; Barbareschi et al. 2003); the CS-PGs brevican (Zhang et al. 1998; Gary et al. 2000; Nutt et al. 2001), versican (Paulus et al. 1996; Ang et al. 1999; Touab et al. 2002) and NG2 (Schrappe et al. 1991; Burg et al. 1997; Chekenya et al. 1999; Shoshan et al. 1999) have also been been implicated.

It would seem that the local balance of PGs in the ECM or pericellular environment is crucial: mechanisms that cause a deficiency (as in arthritic diseases) or abnormal accumulation of HS- and CS-PGs (as in astrocytosis and neurodegenerative lesions) have been associated with damaging pathology. The presence of proteoglycans under normal conditions is clearly necessary and crucial for cellular maintenance: those mechanisms that upset the balance of proteoglycans and their sulphation patterns still need to be elucidated.

Perineuronal nets

Structure and composition

PNNs were first described by Golgi and Cajal in the 1890s as reticular networks observed on the surface of neuronal cell bodies and proximal dendrites. They develop postnatally, finally appearing as net-like features on the cell surface as a result of ECM materials deposited around synaptic endings, and in the space between neurons and astrocytic processes. PNNs were originally visualized using iron- or silver-based histochemical stains; most PNNs can be detected using lectins such as the Wisteria floribunda agglutinin lectin, which recognizes N-acetylgalactosamine in CS-GAG chains (Hartig et al. 1992). Although it is not fully understood exactly which cell types produce the PNN, or why, studies so far have found PNNs most commonly around parvalbumin-containing GABAergic interneurons (e.g. Hartig et al. 1992, 1994; Brückner et al. 1994; Murakami et al. 1995) and pyramidal cells (Hausen et al. 1996) in the cortex, and around projection neurons and large motorneurons of the brain stem and spinal cord (70–80% neurons in the cord have PNNs) (Asher et al. 1995; Murakami et al. 1995; Takahashi-Iwanaga et al. 1998). Because glial cells in vitro can produce ECM similar to PNNs observed in vivo, even in the absence of neurons (Maleski & Hockfield, 1997), it seems likely that astrocytes contribute at least in part to PNN formation. Interestingly, in the culture conditions utilized by Maleski & Hockfield (1997), neurons were never seen to assemble any kind of pericellular matrix – although there is a possibility that this was an artefact resulting from the conditions because Knudson & Knudson (1991) demonstrated that some cells in vitro never produce pericellular matrices, despite having HA receptors and the ability to assemble PNN-like matrix from exogenously added HA and aggrecan.

The structure and composition of PNNs has become more widely understood in recent years. Asher et al. (1995) demonstrated immunohistochemically the existence of major cartilage constituents in the perineuronal matrix of bovine spinal cord: aggrecan and link protein, both of which were hyaluronate-dependent as their immunoreactivities could be abolished by pretreating with the enzyme hyaluronidase. Aggrecan is a large proteoglycan belonging to the lectican family of CS-PGs, which also includes brevican, neurocan and versican. Whereas aggrecan and versican are present in a number of connective tissues (e.g. Mundlos et al. 1991; Yamagata et al. 1993; Glumoff et al. 1994; Popp et al. 2003), brevican and neurocan are usually found only in neural tissues (Rauch et al. 1992; Oohira et al. 1994; Yamada et al. 1994; Seidenbecher et al. 1995; Watanabe et al. 1995). All four of these lecticans have since been detected in PNNs (Bignami et al. 1993; Koppe et al. 1997a, b; Brückner et al. 1998; Matsui et al. 1998; Matthews et al. 2002), as has phosphacan/DSD-1 (Wintergerst et al. 1996; Haunso et al. 1999) and tenascin-R (Brückner et al. 2000; Haunso et al. 2000). In a recent review, Yamaguchi (2000) proposed a model of PNNs whereby long HA molecules bind CS-PGs (lecticans), which in turn bind tenascin and form net-like complexes between neurons and glia.

PNNs are not all identical in composition: whereas PNNs in the rat deep cerebellar nuclei and spinal cord, for example, are rich in expression of phosphacan, simultaneously labelled sections of cortex demonstrate little phosphacan (Fig. 2). Antibodies to the aggrecan-like antigen CAT-301 appear to label PNNs as frequently as the Wisteria floribunda lectin, whereas versican expression is more distinguishable around large motorneurons in the spinal cord and brainstem. The importance of differential expression patterns of CS-PGs in PNNs is unclear; however, tenascin appears to be a more critical component, because mutant tenascin-knockout mice showed disrupted PNNs (Weber et al. 1999; Brückner et al. 2000; Haunso et al. 2000), whereas mice deficient in neurocan (Zhou et al. 2001) or brevican (Brakebusch et al. 2002) were anatomically and morphologically similar to wild-type mice.

Fig. 2.

Immunohistochemistry demonstrating PNNs in 4% paraformadlehyde-perfused adult rat CNS using Wisteria floribunda agglutinin lectin (WFAL; A–C); antibodies to phosphacan (2B49; D–F); and antibodies to versican (12C5; G–I). PNNs are clearly demarcated with WFAL around small neurons in the cortex (A) and larger motorneurons in deep cerebellar nuclei (DCN; B) and spinal cord (C). Phosphacan expression in cortex (D) is faint with few distinguishable PNNs, although strong perineuronal labelling is seen in DCN (E) and spinal cord (F). Strong expression of versican throughout the cortex (G) makes it harder to distinguish individual PNNs, although perineuronal expression of versican appears stronger in DCN (H) and spinal cord (I). Scale bars = 50 µm except (C) 10 µm; arrows point to examples of perineuronal nets.

Functional roles for CS-PGs and PNNs in synaptic plasticity

The late development of PNNs, and their association with only certain neuronal types, has led to speculation of their functions including roles in maintenance of tissue architecture (Margolis & Margolis, 1993; Celio & Blumcke, 1994) and involvement in maturation of synapses on motorneurons (Kalb & Hockfield, 1988, 1990). Following the observation that neuronal surfaces covered by PNNs are devoid of synaptic contacts (Brückner et al. 1993), it has been proposed that PNNs may specifically serve to block formation of new synapses (Celio & Blumcke, 1994; for review see Yamaguchi 2000).

It has been shown that the late postnatal appearance of PNNs in the visual cortex coincides with the end of the so-called critical period (Sur et al. 1988; Zaremba et al. 1989; Koppe et al. 1997a, b), a well-documented developmental time-window during which synapses are still open to experience (e.g. Pettigrew, 1972; Daw & Wyatt, 1976) (for a review of synaptic plasticity and ECM molecules see Dityatev & Schachner, 2003). At the end of the critical period, synaptic plasticity is greatly diminished – however, visually deprived (dark-reared) rats retain plasticity of the visual cortex (Cynader et al. 1976): the phenomenon has been well characterized (recently reviewed by Berardi et al. 2003). Studies from Hockfield and colleagues previously demonstrated changes in expression of CS-PGs in the visual cortex in visually deprived rats (Zaremba et al. 1989; Guimaraes et al. 1990; Hockfield et al. 1990), implying a role for CS-PGs and PNNs in maintaining synaptic stability and preventing plasticity in the mature animal (for review see Hockfield & Kalb, 1993).

In support of these earlier studies, Pizzorusso et al. (2002) demonstrated that treatment of the mature rat visual cortex with chondroitinase ABC degrades CS-PGs present in the PNNs, restoring synaptic plasticity normally only seen during the critical period in the development, or in dark-reared animals. Furthermore, mutant mice lacking neurocan (Zhou et al. 2001) or brevican (Brakebusch et al. 2002) were shown to have defects in hippocampal long-term potentiation although their brains were anatomically and morphologically indistinguishable from those of wild-type mice. Bukalo et al. (2001) found differential effects of removal of tenascin-R (using knockout mice) or CS-PGs (using chondroitinase-ABC) on hippocampal synaptic plasticity. These findings have important implications for the future, showing that it may be possible to restore growth potential and plasticity to the mature CNS.

Conclusions

The mature CNS is highly prohibitive to establishment of new synaptic connections or regeneration of injured neurons. Research over the past few decades has highlighted many molecules (in particular, CS-PGs) that appear to be involved in both maturation of synapses and prevention of new axonal growth. Although it must be assumed that these mechanisms exist in order to prevent aberrant neuronal growth and accidental rewiring of the CNS, recent advances have demonstrated that manipulation of the environment using substances such as chondroitinase-ABC can override these barriers. The findings suggest exciting prospects for the future, particularly for therapeutic research in overcoming debilitating spinal cord injury, and increasing CNS plasticity; however, many questions remain to be answered. Some protective roles for CS-PGs have been highlighted; it is clear that abnormal deficiency or accumulation of CS-PGs leads to pathological states; further investigation of the long-term effects of manipulating the balance of ECM molecules in the CNS needs to be conducted.