Extracellular matrix and perineuronal nets in CNS repair


  • Jessica C.F. Kwok,

    1. Cambridge Centre for Brain Repair, Department of Clinical Neurosciences, University of Cambridge, Cambridge CB2 0PY, United Kingdom
    Search for more papers by this author
  • Gunnar Dick,

    1. Cambridge Centre for Brain Repair, Department of Clinical Neurosciences, University of Cambridge, Cambridge CB2 0PY, United Kingdom
    2. Department of Molecular Biosciences, The Faculty of Mathematics and Natural Sciences, University of Oslo, N-0316 Oslo, Norway
    Search for more papers by this author
  • Difei Wang,

    1. Cambridge Centre for Brain Repair, Department of Clinical Neurosciences, University of Cambridge, Cambridge CB2 0PY, United Kingdom
    Search for more papers by this author
  • James W. Fawcett

    Corresponding author
    1. Cambridge Centre for Brain Repair, Department of Clinical Neurosciences, University of Cambridge, Cambridge CB2 0PY, United Kingdom
    • Cambridge Centre for Brain Repair, Department of Clinical Neurosciences, University of Cambridge, Cambridge CB2 0PY, United Kingdom
    Search for more papers by this author


A perineuronal net (PNN) is a layer of lattice-like matrix which enwraps the surface of the soma and dendrites, and in some cases the axon initial segments, in sub-populations of neurons in the central nervous system (CNS). First reported by Camillo Golgimore than a century ago, the molecular structure and the potential role of this matrix have only been unraveled in the last few decades. PNNs are mainly composed of hyaluronan, chondroitin sulfate proteoglycans, link proteins, and tenascin R. The interactions between these molecules allow the formation of a stable pericellular complex surrounding synapses on the neuronal surface. PNNs appear late in development co-incident with the closure of critical periods for plasticity. They play a direct role in the control of CNS plasticity, and their removal is one way in which plasticity can be re-activated in the adult CNS. In this review, we examine the molecular components and formation of PNNs, their role in maturation andsynaptic plasticity after CNS injury, and the possible mechanisms of PNN action. © 2011 Wiley Periodicals, Inc. Develop Neurobiol 71: 1073-1089, 2011


The perineuronal net (PNN) is a layer of condensed pericellular matrix which aggregates and wraps around the soma and proximal dendrites of some neurons in the central nervous system (CNS). It was first reported by Camillo Golgi at the end of the 19th century and was once regarded as a staining artifact (Golgi,1898). In the last few decades, there have been numerous studies on the molecular composition and anatomy of this structure. The connection between this extracellular matrix (ECM) structure and maturation of the CNS has also been investigated. Enzymatic perturbation of PNNs results in enhanced morphological and functional plasticity in the CNS. However, the molecular mechanism(s) underlying these properties of the PNN are still largely unknown. Recent studies on PNN-binding molecules and chondroitin sulfate proteoglycans (CSPGs) receptors/interacting partners provide a new dimension for the understanding of this structure. In this review, we aim to systematically examine these recent discoveries and understand how PNNs are involved in controlling CNS plasticity after injury.

The Highly Organized Complex of the PNN

The PNN is composed of ECM molecules which are strongly expressed in the nervous system. The main components of PNNs are hyaluronan (HA), link proteins, CSPGs, and tenascin-R (Tn-R) (Köppe et al.,1997; Carulli et al.,2007; Kwok et al.,2010). Through specific interactions, these ECM molecules form large and stable aggregates on the surface of somata and proximal dendrites in sub-populations of neurons [Fig. 1(A)]. CSPGs in the extracellular space interact with the linear, non-sulfated HA polymer chains on the cell surface. This interaction is stabilized by the presence of link proteins, a group of proteins which bind to both HA and CSPGs. The C-terminal domain in the core protein of the CSPGs then binds to the trimeric Tn-R, forming the highly organized structure of PNN (Fig. 1; Lundell et al.,2004). This layer of organized matrix appears as a coat on the neuronal surface, and, as we will explain, controls the formation of synapses and connections which are important for plasticity.

Figure 1.

The structure of the PNN. A: PNNs stained with the use of WFA in rat brains. The staining is mainly observed on the surface of somata and proximal dendrites in neurons. B: HAS, present on the neuronal surface, synthesize HA and secrete it to the perineuronal area. Members of the lectican family, including aggrecan, neurocan, versican, and brevican, then bind to this HA backbone. This binding is consolidated by the presence of link proteins. While aggrecan is present on almost all PNN-positive neurons, other lecticans are found in subpopulations of PNN neurons. The N-terminal domain of lecticans harbors a HA-binding domain, a GAG attachment region is in the middle of the molecules while the C-terminus bears a C-type lectin-like domain. The length of the GAG attachment varies among the family members, with aggrecan being the longest. The C-terminus of the lecticans binds to Tn-R and the N-terminus binds to the HA backbone. This arrangement allows the formation of supramolecular aggregates on the surface of neurons. The G3 domain of the lecticans then binds to the Tn-R trimer to finish the complex structure of PNNs. Scale bar, 50 μm.

Hyaluronan and Hyaluronan Synthases.

HA is a linear polymer of N-acetylglucosamine and glucuronic acid (GlcA) disaccharide units, linked by β1-4 and β1-3 linkages (Meyer et al.,1951) and non-sulfated. It is the only glycosaminoglycan (GAG) which incorporates into the ECM without being covalently attached to a protein core (Toole,2000). Similar to cartilage, HA interacts with other ECM molecules in the nervous system, leading to the formation of large aggregates (Hardingham and Muir,1972; Hascall and Heinegard,1974; Köppe et al.,1997). Electron micrographs of proteoglycans isolated from adult rat brains under non-denaturing conditions showed a typical feather-like structure, resembling the HA–aggrecan complex in cartilage (Iwata et al.,1993). Staining using biotinylated hyaluronan binding protein (HABP), a group of proteins showing a high binding affinity to HA, reveals a clear perineuronal pattern in brain sections while digestion with hyaluronidase completely abolishes this staining (Köppe et al.,1997).

HA is synthesized by a family of enzymes called hyaluronan synthase (HAS) (Table 1). The enzymes are located at the inner surface of the plasma membrane, utilizing both monosaccharides (N-acetylglucosamine and GlcA) alternatively in the synthesis of the HA polymer (Philipson and Schwartz,1984; Weigel et al.,1997; Weigel and DeAngelis,2007). The three isoforms of HAS, HAS-1, -2, and -3, are encoded by individual HAS genes located on different chromosomes. They synthesize HA chains of various lengths and with different speeds. While HAS-3 produces short HA chain at the slowest speed, HAS-1 or -2 produce HA chains of up to 2000 kD with speeds >10 times of that of HAS-3 (Spicer et al.,1997; Itano et al.,1999). With the use of in situ hybridization, it has been shown that PNN neurons in the cerebellum express HAS-2 and HAS-3, while those in the spinal cord express HAS-1 and HAS-3 (Carulli et al.,2006, Galtrey et al.,2008). There are also temporal changes of HAS expression in the spinal cord. While both HAS-1 and -3 are expressed in developing spinal cord (between postnatal day 3–21), HAS-3 is the only isoform identified in the mature spinal cord (Galtrey et al.,2008). These changes in HAS expression could be functionally significant because the different HASs synthesize HA of different lengths at different speeds, and could affect PNN properties. It is possible that the differences in chain lengths resulting from differential HAS expression confer different structure and mechanical strength to the PNNs, similar to the experiments reported in pig heart valves where regions of high tensile strength in heart valves are associated with reduced GAG chain length and content (Stephens et al.,2008).

Table 1. Molecular Characteristics of the Three Hyaluronan Synthases
 Chromosome LocationExpected Chain LengthMolecular Weight (kDa)Thickness of HA CoatStability
HAS-119q13.3-q13.417200–200065ThinLow (∼1 h)
HAS-28q24.1215200–200063ThickModerate (∼4 h)
HAS-316q22.18100–100063ThickHigh (∼8 h)

One key question is how does the PNN attach to the neuronal surface despite the presence of all of the components throughout the brain ECM? Various studies on the transmembrane/membrane-bound PNN components (such as phosphacan/receptor protein tyrosine phosphatase (RPTP)-β/ζ and brevican) and the receptors/binding partners of the PNN components [such as CD44, RHAMM (receptor for hyaluronic acid-mediated motility) and LYVE-1 (lymphatic vessel endothelial HA receptor 1)] have shown that they are not responsible for locating the PNNs onto the cell surface. Recently, there are reports showing that RPTPσ and contactin-1 are major receptors for CSPGs in the nervous system (Mikami et al.,2009; Shen et al.,2009). Although the relationship between these molecules and PNNs has not been investigated, the results from immunostaining in neuronal tissues did not show obvious relation to the PNNs. These results suggest that transmembrane/membrane-anchored proteins or HA-binding molecules are not responsible for anchoring the PNN in the nervous system. With the use of in vitro cell models, our group has recently demonstrated that HASs are in fact the anchors for the PNN. In situ hybridization of HASs in cerebellum and spinal cord has shown that these enzymes are expressed in almost all PNN-bearing neurons (Carulli et al.,2007; Galtrey et al.,2008). Expression of HAS-3, together with the presence of soluble link proteins, in human embryonic kidney cells allowed the formation of a PNN-like matrix on the cell surface. This matrix is positive for Wisteria floribunda lectin (WFA) staining and is resistant to stringent detergent and urea washes, similar to the endogenous PNN in brain sections (Deepa et al.,2006). As proposed by Spicer et al. (1997), HAS has to perform at least six different enzymatic/binding activities in order to synthesize a HA chain from the N-acetylglucosamine and GlcA subunits: binding for the two sugar monosaccharides, two different glycosyltransferase activities, a ratchet-like transfer reaction that moves the growing polymer one sugar at a time, and binding site(s) that anchor the growing HA polymer to the enzyme (Spicer et al.,1997; Spicer and Tien,2004). Our results from the in vitro models provide a strong evidence for the last activity of HAS: that neuronal surface-associated HAS anchors the HA chains to the neuronal surface and thus allows for the aggregation and formation of the PNN (Kwok et al.,2010).


Lecticans, also called hyalectans, are members of the CSPG family which possess the ability to bind to both HA and lectins (Ruoslahti,1996; Iozzo,1998). Linear and unbranched sulfated chondroitin sulfate (CS) GAG chains are covalently attached to the lectican protein core through serine residues to form CSPGs. There are four members of lecticans—aggrecan, versican, neurocan, and brevican. While aggrecan, versican, and neurocan are mostly found in the ECM, brevican is found either in the ECM or linked to the cell membrane via a glycosylphosphatidylinositol anchor (Seidenbecher et al.,1998). Reports have shown the presence of all four lecticans in the PNNs (Matsui et al.,1998; Hagihara et al.,1999; Carulli et al.,2006). While aggrecan is present on almost all PNN-positive neurons, other lecticans are only found in subpopulations of PNN neurons (Galtrey et al.,2008). Aggrecan is necessary for PNN formation in our in vitro HEK cell model, and organotypic cultures derived from aggrecan knockout animals fail to form normal PNNs (Giamanco et al.,2010).

Figure 2 shows the structures of lecticans. All lecticans share a similar molecular structure: (1) an N-terminal G1 domain consisting of an immunoglobulin (Ig) repeat and two link modules, (2) the middle GAG attachment region, and (3) a C-terminal G3 domain consisting of two EGF-like repeats, a C-type lectin-like domain and a complement regulatory protein-like module (Iozzo,1998). In aggrecan, there is an additional G2 domain, consisting of two link modules, located immediately after the G1 domain at the N-terminus. While the G1 domain is important for binding to HA chains, the C-type lectin domain in the G3 domain is responsible for binding to Tn-R glycoproteins (Aspberg et al.,1997; Yamaguchi,2000; Lundell et al.,2004). The length of the middle GAG attachment domain varies greatly among the family members. This variation determines the final sizes of the core proteins, ranging from ∼90 kD in brevican to ∼400 kD in versican (Ruoslahti,1996; Zimmermann and Dours-Zimmermann,2008). The interaction between HA and lecticans is stabilized by a family of molecules called link proteins (as discussed below). This arrangement allows the formation of supramolecular aggregates on the surface of neurons. As with HA, the properties of lecticans may be crucial to the stability and compactness of the PNN. A lectican (e.g., brevican) with higher affinity for Tn-R might form a more highly cross-linked PNN than those with lower affinity (e.g., neurocan) (Aspberg et al.,1997; Yamaguchi,2000). The sizes of the lectican molecules and the glycosylations in the middle GAG attachment regions may also affect the degree of cross-linking in the PNN and therefore affect its compactness. This principle has been demonstrated in arterial smooth muscle pericellular matrix, where the overall thickness of the pericellular matrix was reduced when the surface HA was attached to the V3 isoform of versican than other isoforms (Evanko et al.,2001). In fact, immunohistochemistry of Wisteria floribunda agglutinin (WFA; a lectin which labels the N-acetylgalactosamine residue on the CS in the PNN) with antibodies against different CSPG core proteins in brain and spinal cord sections showed that different PNNs can contain different CSPGs (Galtrey et al.,2008). Staining with aggrecan antibodies recognizing different glycosylation variants also shows a differential distribution (Matthews et al.,2002). This suggests a differential distribution of various core proteins on individual PNN positive neurons, which might modulate the properties of the PNN, and therefore affect PNN functions.

Link Proteins.

Link proteins are a group of proteins that interact with both HA and CSPGs and thus stabilize the HA-CSPG aggregates. They belong to the link module superfamily, which includes link proteins and hyalectans. Based on their properties in binding to HA and CSPGs, they also belong to the hyaluronan- and proteoglycan-binding link protein gene family (HAPLN) (Spicer et al.,2003). There are four members of HAPLN, with cartilage link protein (Crtl1 or HAPLN1) being the first member identified (Table 2; Hascall and Heinegård1974; Neame et al.,1986; Marks et al.,1990). All four link proteins are around 38–43 kDa and share a similar structure: an N-terminal signal sequence, an Ig domain, and two consecutive link modules (or proteoglycan tandem repeats), which are important for binding to HA [Fig. 1(B); Hirakawa et al.,2000]. Sequence analysis of the amino acids among members shows that the overall similarities are ∼52–62%, with the N-terminal and the Ig domains sharing the least similarities (Spicer et al.,2003).

Table 2. The Different Nomenclature of the Link Protein Family Members
 Other NamesExpression in TissueJuxtaposed Genes
HAPLN 1Cartilage link protein-1 (Crtl-1)Cartilage/CNSVersican
HAPLN 2Brain link protein (Bral-1)CNSBrevican
HAPLN 3 Widely expressed in different tissuesAggrecan
HAPLN 4Brain link protein-2 (Bral-2)CNSNeurocan

To date, three of the four link proteins have been found in the nervous system; HAPLN3 is the only member not expressed in the nervous system. Northern blot analysis showed that HAPLN3 is mostly expressed by smooth muscle and placenta (Tsifrina et al.,1999). Brain link protein (Bral)-1 (or HAPLN2) is found exclusively in the adult nervous system by Northern blot analysis. Immunostaining showed that Bral1 is mainly associated with myelinated axons in the brain and spinal cord of adult animals and it colocalizes with versican at the nodes of Ranvier to form PNN-like structure, the perinodal net. A recent study in Bral1 knockout mice showed abnormal diffuseness of HA-associated ECM in the node and a marked decrease in conduction velocity in the spinal cord (Hirakawa et al.,2000; Oohashi et al.,2002; Bekku et al.,2010). Crtl1 and Bral-2 are the members which co-localize with PNN. In situ hybridization demonstrated that they are expressed exclusively by PNN-bearing neurons (Bekku et al.,2003; Rauch et al.,2004; Carulli et al.,2006; Galtrey et al.,2008). With the use of our in vitro model for examining the assembly and structure of the PNN, we showed that absence of Crtl1 in PNN-bearing cells prevents the formation of a compact surface pericellular matrix (Kwok et al.,2010). This result suggests that the presence of link proteins in the PNN is crucial for the condensed nature of the PNN. This is also confirmed in animals that lack Crtl1in the CNS. Knockout mice exhibit reduced and attenuated PNNs throughout the nervous system with a complete absence of dendritic WFA staining, and faint and diffuse staining on neuronal somata (Carulli et al.,2010).


Tenascin (Tn) belongs to a family of modular glycoproteins, consisting of five members, Tn-C, -R, -W, -X, and -Y. They are multimeric ECM molecules important in morphogenesis and neural development, nerve regeneration, cell migration, and wound healing. All tenascins contain an N-terminal Tn-assembly domain, a heptad domain, EGF-like repeats, fibronectin III repeats, and a fibrinogen-like globule (Jones and Jones,2000). While Tn-C and -X are widely expressed outside the nervous system (such as dense connective tissue and smooth muscle), the expression of Tn-R and Tn-W is more restricted, with Tn-W mainly in bone and Tn-R in the nervous system (Bristow et al.,1993; Kimura et al.,2007; Tucker and Chiquet-Ehrismann,2009). The heptad domain in theN-terminal is responsible for the multimerization of Tn, resulting in either a homo-trimer in Tn-R [Fig. 1(B)] or a homo-hexamer in Tn-C and -W.

Immunostaining of Tn-R showed co-localization with WFA-staining in PNNs (Weber et al.,1999; Carulli et al.,2006). The G3 domain of lecticans binds to the fibronectin III repeats in the Tn-R trimers in a calcium-dependent manner [Fig. 1(B); Lundell et al.,2004; Zimmermann and Dours-Zimmermann,2008]. As one Tn-R trimer can bind up to three lectican molecules, this Tn-R–lectican interaction helps in strengthening the macromolecular nature of the PNN. Tn-R knockout mice show abnormal PNN staining, with less regular distribution of WFA staining around the perikarya and also less punctuate staining in the dendritic shafts (Weber et al.,1999).

The Role of the PNN in Neurons

The PNN has been observed surrounding the neurons throughout the CNS, including visual cortex, barrel cortex, deep cerebellar nuclei, substantia nigra, hippocampus, and spinal cord (Brauer et al.,1993; Morris and Henderson2000; Pizzorusso et al.,2002; Carulli et al2006; McRae et al.,2007; Brückner et al.,2008; Galtrey et al.,2008). It has also been identified in different organisms, such as mouse, rat, opossum, monkey, and human (Brückner et al.,1998b; Sayed et al.,2002; Horn et al.,2003; Hilbig et al.,2007; Schmidt et al.,2010).With the use of WFA staining, the PNN has mainly been localized to parvalbumin (PV)-positive GABAergic interneurons in the CNS, although it has also been described in other neurons, such as pyramidal neurons from parietal cortex and neurons in medial nucleus of the trapezoid body (Takahashi-Iwanaga et al.,1998; Härtig et al.,2001; Wegner et al.,2003). Aggrecan antibodies give a slightly different distribution, with PNNs around neurons in most layers of the cortex, with different neurons surrounded by aggrecan with different glycosylation patterns (Matthews et al.,2002).

There are a number of hypotheses about the function of PNNs, and they may have several functions. However, it is now clear that they are centrally involved in the control of CNS plasticity. The PNN has been repeatedly observed ensheathing highly active neurons, suggesting that it supports the functions of these neurons (Härtig et al.,1994,1999,2001; Wintergerst et al.,1996; Morris and Henderson,2000; Reimers et al.,2007). In the medial septal/diagonal band complex of rat, triple immunofluorescence combined with intracellular recording showed PNNs on fast-spiking neurons with narrow action potentials, but not on slow-firing neurons, burst-firing neurons or regular spiking neurons (Morris and Henderson,2000). Similar observations have also been reported in rat neocortex and visual cortex (Wintergerst et al.,1996; Reimers et al.,2007). It has therefore been proposed that the PNN provides a suitable micro-environment around these neurons for their high activity, possibly to its buffering capacity of local cations. Due to the highly negative charge of GAGs, the PNN is capable in binding to various cations in the local environment, controlling the diffusion of ions such as sodium, potassium, and calcium, thus serving as a rapid cation exchanger to support the high activity of these neurons, and they could also create extra space for extracellular fluid (Härtig et al.,1999,2001). However, treatment with chondroitinase ABC (ChaseABC) does not cause epileptiform activity, neither does it change baseline activity levels, as might be predicted from this hypothesis (Pizzorusso et al.,2002). CSs have been shown to bind extracellular calcium in physiological concentration and digestion of CS with ChaseABC increases the rate of calcium diffusion in both neocortex and hippocampal slice culture (Hrabetová et al.,2009). With the use of patch-clamp recording in Xenopus photoreceptors, Vigetti et al. showed that both chondroitin 4- and 6-sulfates (also called CS-A and -C, the two most common isoforms of CS disaccharides; Fig. 2) are able to shift the activation curve of voltage-dependent calcium channels when the photoreceptors were perfused with CS solutions under physiological extracellular calcium concentration. This suggests that the CSs are either binding to the extracellular calcium and therefore control its availability to the receptors or that the CSs have a direct interaction with the calcium channels (Vigetti et al.,2008). In the brain regions controlling monkey and human saccadic movement systems, Horn et al. have shown that PNNs are present in two groups of fast-spiking neurons (mesencephalic reticular formation and paramedian pontine reticular formation in the midbrain). The saccadic omnipause and burst neurons are both fast-spiking neurons. While saccadic omnipause neurons are glycinergic inhibitory neurons, saccadic burst neurons are mainly glutamatergic or aspartatergic excitatory neurons. Taken together these observations suggest that PNNs may be involved in the function of fast-spiking neurons regardless of their transmitter or postsynaptic action (Horn et al.,2003).

The main function of PNNs appears to be in synapse formation synaptic plasticity. A recent study in a co-culture system of hippocampal neurons with primary astrocytes showed that the development of synapses occurs simultaneously with the emergence of the PNN. Enzymatic digestion of the ECM components with ChaseABC or hyaluronidase increased the number of synaptic puncta in the neurons. The neurons, however, showed a decreased sensitivity to glutamate application suggesting the role of the CS/HA in controlling the receptor density at the synapses (Pyka et al.,2011). A direct effect of PNNs on receptor distribution was recently shown with the use of single-molecule tracking technology and time-lapse imaging. Frischknecht et al. demonstrated that the diffusion constant of the AMPA receptor subunit GluR1 drops significantly as soon as it approaches an HA-rich area on the cell surface. Removal of the ECM in the PNN with hyaluronidase digestion increased both the diffusion coefficient and the total surface area explored by single subunits by 20–150% (Frischknecht et al.,2009). Readers can refer to other articles in this issue by Wlodarczyk et al. (2011) or Dansie and Ethell (2011) for more details on the effect of the ECM on synaptic plasticity.

Activity-Dependent Regulation of PNN Formation

The formation of the PNN roughly coincides with the maturation of the nervous system (Pizzorusso et al.,2002; Carulli et al.,2006; Galtrey et al.,2008) suggesting that its formation and maturation may depend on neuronal activity. Experiments using a nonselective blocker of voltage-gated sodium channels, but not transgenic models with compromised glutamate release, clearly inhibited PNN formation (Reimers et al.,2007). Similar observations have also been reported in the visual cortex and barrel cortices (Dityatev et al.,2007; McRae et al.,2007). Dark rearing in newborn cats prolongs the critical period in the visual cortex with a concomitant decrease in aggrecan immunostaining (Hockfield et al.,1990; Lander et al.,1997; Pizzurusso et al.,2002), but in developing chicks light deprivation did not prevent PNN formation (Gati et al.,2010). The critical period refers to a stage during development where appropriate experience drives the organization of a neuronal network and absence of correct experience may lead to formation of incorrect neural connections. In the mouse barrel cortex, PNNs are expressed in an activity-dependent manner. Manipulating sensory input through continuous whisker trimming from postnatal day (P) 0 to P30 leads to reduced numbers of aggrecan-positive PNNs in layer V of barrel cortex. Whisker trimming after P30 does not have this effect. Interestingly, after 30 days of normal activity post-whisker trimming, there is a further reduction in aggrecan-positive PNN levels, implying that experience in first 30 days of life is critical. The decrease appears to be specific to aggrecan, but not other PNN markers, localized specifically around PV-expressing cells (McRae et al.,2007). Sensory deprivation of facial vibrissae in rat barrel cortex also reduced the number of WFA-labeled PNNs (Nakamura et al.,2009). However, in rodent visual system light deprivation had its greatest effect on expression of the link protein Crtl1 (Carulli et al.,2010). Taken together, the data suggest that there may be an important role for aggrecan and for link protein in the activity-dependent formation of PNNs and for the expression of PNNs in regulating the close of the critical period.

ECM and PNNs in CNS Injuries

PNNs are dynamic structures that can undergo extensive changes after injury and changes in activity, including changes to their CSPG core proteins and sulfation. Figure 2 shows the most common sulfation isoforms of CS. A consequence of CNS injury in humans and other mammals is the formation of a glial scar, resulting from the proliferation of astrocytes, oligodendrocyte progenitor cells, meningeal fibroblasts, and also pericytes, migrating into the lesion (Hatten et al.,1991; Chen et al.,2002; Göritz et al.,2011). There is a general up-regulation of proteoglycans, particularly CSPGs, within the glial scar. The increased CSPGs remain in the injured sites for a long period of time after injury (Fawcett and Asher,1999; Silver and Miller,2004). More specifically, numerous studies have shown increases in neurocan, versican, and brevican after CNS injury in both human and other mammals (Jones et al.,2002,2003; Buss et al.,2009). Changes to aggrecan, however, are less clear. There are reports showing a decrease of aggrecan after hemisection injury in the spinal cord (Lemons et al.,2001,2003). However, there is also a reported decrease of CSPGs (aggrecan, neurocan, and phosphacan) within the developing glial scar after traumatic brain injury, alongside a significant increase of versican mRNA around the lesion site (Harris et al.,2009). Among many possibilities, the different proteoglycan changes could be due to the nature and location of the lesion. In a post-mortem study of human spinal cord injury (SCI), phosphacan and NG2 were found to be associated with the glial scar while neurocan and versican were detected within the lesion epicenter associating with invading Schwann cells (Buss et al.,2009). ECM changes have also been observed further away from the lesion site. Using real time PCR and in situ hybridization, Waselle et al. reported different patterns of up-regulation after a dorsal root injury: in the dorsal root entry zone, there are sharp increases in brevican, neurocan, and versican (V1 and V2), but in the dorsal column, the increases in neurocan and brevican are slower and less prominent (Waselle et al.,2009). After cervical dorsal root lesion, neurocan, brevican, and NG2 are up-regulated in dorsal column nuclei by reactive glia cells such as microglia and astrocytes, creating a barrier to regeneration in these uninjured nuclei (Massey et al.,2008).

Figure 2.

A schematic diagram showing the structures of lecticans. Aggrecan, versican, neurocan, and brevican are members of the lecticans. All lecticans contain an N-terminal G1 domain which consists of an Ig repeat and two link modules, a middle GAG attachment region and a C-terminal G3 domain which consists of two EGF-like repeats, a C-type lectin-like domain and a complement regulatory protein-like module. Aggrecan contains an additional G2 domain whose function is not clear. While the G1 domain is important for HA binding, the C-type lectin domain in the G3 domain is responsible for binding to the glycoprotein Tn-R. The length of the middle GAG attachment domain is highly variable among lecticans and it determines the final sizes of the core proteins. The middle box shows the attachment of a CS-GAG chain to the serine residue on a core protein, via a tetrasaccharide linkage. CS is a linear polymer of GlcA and GalNAc subunits. ChaseABC digests the CS chain into its basic disaccharide units, hence removing the GAG chain from the core protein. The disaccharide subunit of CS can be modified with different sulfations. CS-A and -C contain either a single C-4 sulfation or a C-6 sulfation on the GalNAc, CS-D contains a C-2 sulfation on the GlcA and a C-6 sulfation on the GalNAc, and CS-E contains both C-4 and C-6 on the GalNAc simultaneously.

In addition to the general up-regulation of various CSPGs after injury, there are also selective changes in the sulfation pattern on the CS residues. The basic disaccharide subunit of CS is composed of GlcA and N-acetylgalactosamine (GalNAc). Sulfation of CS takes place on carbon (C-) 2 or 3 on the GlcA residue and C-4 or C-6 on the GalNAc residue. A combination of one or more sulfations at various locations on the two sugar subunits gives rise to the different CS isoforms. Figure 2 shows the nomenclature of the common CS isoforms. Chondroitin 4-sulfate is the most prominent CS-GAG found in an adult CNS and also after injury. Other reports have also demonstrated up-regulations of 4,6-sulfate, 6-sulfate, and 2-sulfate in injured brain (Gilbert et al.,2005; Properzi et al.,2005; Wang et al.,2008).

CS-GAGs are generally considered to be inhibitory to neural regeneration after injury, as demonstrated by numerous ChaseABC digestion studies (see below). However, recent studies on CS disaccharides suggest that some CS isoforms may in fact be beneficial to neural regeneration. Rolls et al. have demonstrated that CS disaccharides resulting from ChaseABC digestion induce neurite outgrowth and protect neurons from neuronal toxicity and axonal collapse in vitro (Rolls et al.,2004). Culture of mouse cerebellar granule neurons on different CS GAGs showed that chondroitin-4-sulfate (CS-A), but not chondroitin-6-sulfate (CS-C), exhibits a strong negative guidance on the neurons (Wang et al.,2008), while another study identified disulfated CS-E as the most inhibitory form (Gilbert et al.,2005). Our recent study in chondroitin 6-sulfotransferase (a key enzyme in adding the sulfate to C-6 on the GalNAc residue of CS) knockout mice showed less regeneration in the CNS when compared to the wild type mice (Lin et al.,2011), suggesting that CS-C may actually be relatively permissive for regeneration. These results indicate that sulfations on the CS GAGs could also regulate the regeneration response after injury.

Manipulations of the ECM in the PNNs and Glial Scar After Spinal Cord Injuries.

In general, CSPGs up-regulation after CNS injury creates a barrier to regeneration as well as restricting plasticity. Reactive astrocytes and oligodendrocyte progenitor cells up-regulate CSPGs and exert a CS-GAG dependent inhibitory effect on the axon growth of dorsal root ganglion (DRG) neurons (Smith-Thomas et al.,1995; Asher et al.,2000). One easy and convenient way of manipulating the ECM in both the glial scar and PNN in the CNS is by application of ChaseABC. ChaseABC is a bacterial enzyme isolated from Proteus vulgaris (Yamagata et al.,1968). It digests CS A, B, and C by an elimination reaction, yielding Δ4,5-unsaturated disaccharides. A large number of studies have shown that the removal of CS chains in the CSPGs renders the CSPGs less inhibitory (Sango et al.,2003; Nakamae et al.,2009). This was first demonstrated in DRG culture. The digestion of CS GAGs removes the inhibitory effect on DRG growth in vitro (McKeon et al.,1995). In the adult CNS, ChaseABC injection removed the WFA staining of the PNNs which could take months to return (Brückner et al.,1998a). After ChaseABC treatment, cryosections from injured adult rat spinal cord support better regeneration of DRG axons (Zuo et al.,1998). These studies have provided a rational foundation for the use of this enzyme in vivo.

The first in vivo injection of ChaseABC demonstrated that its injection into injured spinal cord degraded CSPGs in the glial scar after contusion injury, implying that ChaseABC digestion is effective in vivo (Lemons et al.,1999). Subsequently, the benefits of ChaseABC on regeneration have been widely observed in different models. Our group has demonstrated that the removal of CSs by ChaseABC promotes dopaminergic axon regeneration in the nigrostriatal tract after injury (Moon et al.,2001). When applied to an acute cervical dorsal column crush lesion, ChaseABC enhances axonal regeneration and locomotor and proprioceptive recovery in these animals, demonstrating that in vivo ChaseABC delivery can lead to anatomical and functional recovery (Bradbury et al.,2002). Regeneration of sensory axons and also axons in Clarke's nucleus after hemisection injury was also observed following ChaseABC treatment (Yick et al.,2003; Shields et al.,2008). Acute and delayed intrathecal ChaseABC infusion to a rubospinal tract lesion reduced the atrophy of neuronal cell bodies of the rubospinal neurons, suggesting a neuroprotective role for ChaseABC (Carter et al.,2011).

Apart from promoting regeneration in injured axons after ChaseABC injection in vivo, extensive evidence also shows that ChaseABC promotes anatomical plasticity in injured CNS, mainly by inducing sprouting of damaged axons and dendrites. When combined with brain-derived neurotrophic factor (BDNF) infusion, ChaseABC promotes sprouting of undamaged retinal axons into the denervated superior colliculus after a partial retinal injury (Tropea et al.,2003). Sprouting of serotonergic, corticospinal, and sensory afferents has been observed in areas deprived of CSPGs after ChaseABC injection in SCI (Barritt et al.,2006; Garcia-Alias et al.,2008). The functional role of these sprouting fibers has been investigated in several injury models. ChaseABC injection in a cervical dorsolateral laceration injury model induced functional collateral sprouting of sensory fibers into the partially denervated cuneate nuclei, and electrophysiological studies of receptive fields in the cuneate nucleus then revealed a significant increase in the areas responsive to forelimb stimulation with ChaseABC treatment, indicating that these sprouting sensory afferents made functional connections (Massey et al.,2006). After dorsal rhizotomy of cervical level (C) 5 to thoracic level 1 which spared C7, ChaseABC injection causes sprouting of C7 primary afferents, restoring post-synaptic responses to C7 dorsal root stimulation (Cafferty et al.,2008). Additional evidence of increased plasticity of spinal circuitry with ChaseABC has been demonstrated by a study in peripheral nerve injury by Galtrey et al. The ulnar and median nerves of the forelimb were cut in adult rat and the severed nerve stumps were surgically sutured back, either by correctly or with cross-over (i.e., the ulnar nerve stump was sutured up with the median nerve stump). The animals were allowed to recover for 4 weeks, followed by a single injection of ChaseABC into the ventral horn of the spinal cord. Subsequent behavioral tests showed that the functional impairments resulted from inaccurate re-innervation in the affected forelimbs could be overcome by a single injection of ChaseABC into the spinal cord. ChaseABC injection led to reduced CSPG, removal of the PNNs around the spinal motor neuron pool, increased sprouting, and recovery of forelimb behavior. This experiment suggests that changes to the ECM and PNNs within the spinal cord allow adaptation of the spinal circuitry to peripheral stimulation and promote functional recovery (Galtrey et al.,2007). Plasticity, when it is available, is driven by environmental or other neurological events. This is also true of ChaseABC-induced plasticity. Thus, treating the cord with ChaseABC after injury only produces a modest improvement in skilled paw function, but the combination of ChaseABC and skilled paw rehabilitation acted together to allow substantial recover (Garcia-Alias et al.,2009). Together these studies suggest that CSPGs inhibit potential plastic changes within the spinal cord and that modification with ChaseABC renders the spinal cord more plastic which allows novel circuit formation. More importantly, these novel connections may transmit physiological information successfully, leading to a meaningful functional recovery.

Functional recovery following ChaseABC treatment has also been demonstrated in a variety of contusion SCI models, which are potentially more clinically relevant. Locomotor and bladder function recovery were demonstrated with intrathecal ChaseABC treatment in adult rats after moderate or severe clip compression injuries (Caggiano et al.,2005). Although when cortical spinal tract regeneration was compared in hemisected and contused models, ChaseABC injection only enhanced the regeneration of cortical spinal tract axons around the injury site in the hemisection model, but not in the contusion model (Iseda et al.,2008). These mixed results suggest that the effects of ChaseABC in sprouting, axonal regeneration, and functional recovery are complicated, and that the lesion parameters such as severity, number of spared axons, location, will be crucial to the recovery in response to ChaseABC.

As well as studies into rodent SCI, ChaseABC has also been applied to SCI models in other mammals, such as cats. After thoracic spinal cord hemisection, repeated intraspinal ChaseABC injections in cats induced in vivo digestion of CSPGs, accelerated recovery of skilled locomotor behaviour (ladder walking and peg-board stepping), improved hindlimb kinematic motion, and increased serotonergic immunoreactivity at the rostral borders of the lesion. This is the first study examining ChaseABC as a potential therapeutic for SCI in larger mammals (Tester and Howland,2008; Jefferson et al.,2011).

Apart from direct injections of ChaseABC to remove CSPGs, alternative strategies have been exploited to deliver the ChaseABC or modify the ECM and PNN in the injury areas. Transgenic animals expressing ChaseABC in astrocytes under a GFAP promoter showed increases in both corticospinal tract regeneration into the injury site after spinal hemisection and in sensory projections across the dorsal root entry zone after dorsal root crush injury (Cafferty et al.,2008). Neutralization of CSPGs with antibodies (e.g., NG2 antibody) or competitive inhibition with the use of oligosaccharides infusion has also been demonstrated to be effective and beneficial to recovery (Tan et al.,2006). Levels of CSPGs can also be reduced by inhibiting CS GAG chain formation. One such method is knocking down the level of chondroitin polymerization factor (ChPF), an enzyme critical in elongation of CS GAG chains. One proof-of-principle study showed reduction in the level of CS produced by astrocytes and Neu7 cells after ChPF siRNA treatment. In addition, conditional medium from Neu7 cells treated with siRNA promoted neurite outgrowth from cerebellar granule neurons (Laabs et al.,2007). Attachment of CS chains to CSPGs has also been prevented by using a DNA enzyme targeted at xylosyltransferase, again with enhanced axon regeneration (Grimpe and Silver,2004).

Other molecules shown to affect CSPGs in the CNS include the small leucine-rich proteoglycan decorin. It has been shown to reduce astrogliosis and basal lamina formation in acute cerebral stab injuries (Yamaguchi et al.,1990; Hocking et al.,1998; Logan et al.,1999; Santra et al.,2002). Infusion of decorin after rat SCI reduced astrogliosis, microglia infiltration, and levels of neurocan, brevican, phosphacan around the lesion. This promotes axonal growth from microtransplanted sensory neurons across the lesion site (Davies et al.,2004).

Manipulation of the PNN in Other Systems in the CNS

The relationship between the PNNs, plasticity, and critical period has been observed elsewhere in the CNS. Dense extracellular PNN surrounds sub-populations of neurons at the time when synaptic connections are stabilized and the critical period ends, such as those from the visual cortex and barrel cortex (Hockfield et al.,1990; Lander et al.,1997; Pizzorusso et al.,2002).

The PNNs in the Visual System.

Upon the closure of a critical period, there is a general up-regulation of CSPG core proteins and changes in GAG sulfation patterns in the visual cortex (Pizzorusso et al.,2002; Carulli et al.,2010). In many parts of the CNS, after the critical period ends, plasticity driven by altered experience is greatly attenuated. In the visual system, if one eye is closed during the critical period, there will be an ocular dominance shift in which both cortices will become more responsive to the remaining open eye. This ocular dominance shift will not happen if monocular deprivation is started after the end of critical period. Removal of CS-GAGs with ChaseABC in adult animals re-opens the critical period as demonstrated by reactivation of ocular dominance plasticity after monocular deprivation (Pizzorusso et al.,2002), implying that ChaseABC can reactivate the plastic state of CNS. For rats which experienced monocular deprivation during the critical period, minimal recovery of the cortical response to the deprived eye is expected in adulthood. However, if ChaseABC was combined with reverse lid suturing where the previously non-deprived eye was sutured to enforce the use of the previously deprived eye at the time of ChaseABC injection, there was complete recovery of visual acuity and dendritic spine density (Pizzorisso et al.,2006), implying that ChaseABC can synergize with training to create an environment where training can exert its effects.

This phenomenon is not restricted to rats. The existence of visual system critical period has been recognized in many mammals including mice, cats, and humans, where early life exposure to visual stimuli is considered critical for proper visual system development. A study in the cat visual system demonstrated a reduction in the number of aggrecan-positive PNNs in adult cats that had been made strabismic and amblyopic as kittens (i.e., by cutting the tendon of the lateral rectus muscle to one eye and the blocking of optic nerves) (Yin et al.,2006). This study demonstrated changes in the visual system of cats relating to activity both before and after the end of the critical period, suggesting activity-dependent changes to PNNs and plasticity is possible even in adult animals. This result has been confirmed in rats; where adult amblyopic rats with long-term monocular deprivation during the critical period recovered visual acuity and cortical responses to the deprived eye after reverse suturing and environmental enrichment. This was accompanied by reductions in PNN numbers and cortical inhibition and an increase in BDNF level in the visual cortex (Sale et al.,2007).

Overall, the studies in the visual system are consistent with the hypothesis that the PNNs restrict plasticity during development. They also demonstrate an interplay between sensory experience and PNN formation and maintenance. Appropriate sensory stimuli are important in triggering formation of PNNs, which in turn, act as a brake in the CNS to stop further changes in response to changing sensory environment. Furthermore, PNNs do not appear to be static structures, but remain sensitive to changes in experience in adulthood, although at much reduced level.

The PNN in Fear Memory.

The relationship between the PNN and the critical period exists elsewhere in the CNS. The PNN is critical in the permanence of fear memory. In adult animals, fear conditioning by pairing an ambient signal with a painful signal produces a permanent memory that may be temporarily inhibited by extinction training, but ultimately is resilient to erasure. In contrast, during early postnatal development, seven days of extinction training after fear conditioning induces the erasure of the conditioned memory; the strength of memory is also reduced in these animals without extinction training, implying active protection of fear memories in adults (Gogolla et al.,2009). Once again the formation of PNNs coincided with the switch in fear memory protection. ChaseABC digestion of PNNs in basalateral amygdala of adult mice allows the subsequent erasure of fear memory by extinction training, suggesting that PNNs are important for the formation of long-term stable fear memories in adults.

The PNN in Hippocampal Synaptic Plasticity.

CSPGs are also involved in functional plasticity of the CNS. ChaseABC treatment reduced long-term potentiation and long-term depression in hippocampal slice cultures (Bukalo et al.,2001; Saghatelyan et al.,2001). Long-term potentiation was abolished in brevican deficient mice and also after injection of anti-brevican antibodies (Brakebusch et al.,2002).

The PNN in Cerebellar Morphological Plasticity.

Exposure of adult mice to an enriched environment for one month causes significant morphological changes such as an increase in the size of axon terminals of Purkinje cells in the cerebellum and reduction in the number of PNNs. These changes are also seen in Crtl1 knockout mice. Environmental enrichment and Crtl1 knockout reduced the level of aggrecan, Crtl1, and HAS-2 in the cerebellar nuclei. Denervation of the deep cerebellar nuclei decreased the intensity of PNN structures (Foscarin et al.,2011). These observations show that plasticity restricting PNNs are regulated by a dynamic interplay between synaptic activities and synthesis and removal of matrix components.

Possible Mechanisms of PNN on Regeneration and Plasticity

A major effect of the PNNs is probably to restrict the formation of new neuronal contacts. As described above, increased sprouting and re-innervation after degradation of the established PNNs have been demonstrated in several injury models (Barrit et al.,2006; Massey et al.,2006; Galtrey and Fawcett,2007). ChaseABC compromises the net structure by degrading CS-GAGs in the PNN, allowing the formation of new synapses, spontaneous sprouting, and increased plasticity (Corvetti and Rossi.,2005; Barrit et al.,2006; Massey et al.,2006; Wang et al.,2011). The efficiency of ChaseABC treatment implicates CS GAGs in the PNN as inhibitors of neuronal plasticity. The molecular mechanisms underlying these inhibitory effects of CS GAGs may be (1) a direct interaction through a CS GAG receptor, (2) a mechanism where CS-GAGs make the PNNs a non-permissive substrate for integrins or other proteins in the adhesion and growth machinery, or (3) an indirect interaction where CS GAGs interacts with and presents growth inhibitory proteins. Recently, contactin-1 (Mikami et al.,2009) and RPTPσ (Shen et al.,2009) have been identified as receptors interacting with CS. The transmembrane receptor, RPTPσ, binds with high affinity to neural CSPGs such as neurocan and aggrecan. The role of RPTPσ in inhibiting neuronal outgrowth is convincingly demonstrated by robust regeneration in the RPTPσ (−/−) knockout after SCI (Shen et al.,2009; Fry et al.,2010). How the GAG moiety might regulate the biological effect of outgrowth is not established. In Drosophila, the RPTPσ homolog—Lar, exerts opposing effects in synaptic development after binding to two different HSPGs, syndecan, and Dallylike (Johnson et al.,2006). As RPTPσ is able to interact differently with HS as well as CS-GAGs, it is possible that the interacting GAG determines the biological effect of RPTPσ on outgrowth (Coles et al.,2011). On the other hand, there are also suggestions that the degree and pattern of sulfation is important in determining outgrowth after injury (Gilbert et al.,2005; Wang et al.,2008). Contactin-1, another recently identified CS receptor, interacts specifically with di-sulfated CS-E. The effect of this interaction is, however, mainly associated with stimulation of neurite outgrowth (Mikami et al.,2009).

A second potential mechanism of repulsion may be through control of integrin activation. Integrins mediate neuronal adhesion and growth, but their growth-promoting activity is highly dependent on the substrate on which the neurons are attached. Neuronal cells growing on substrate coated with CSPGs, such as aggrecan, demonstrated outgrowth inhibition accompanied by reduced integrin activation (Afshari et al.,2010). The inhibition was overcome by integrin up-regulation or by increased integrin activation using activating antibodies (Condic et al.,1999; Afshari et al.,2010; Tan et al.,2011). In some systems, the negative effects of CSPGs on integrin activation and substrate interaction have been reduced by ChaseABC (Afshari et al.,2010), which implies that the CS GAGs modulate integrin activation. However, direct interaction between integrins and CS GAGs has not been reported. Alternatively, removal of CS GAGs, may unmask integrin-activating domains in the CSPG protein cores or in other matrix proteins. More information on ECM and integrin interactions can be found in the other review by Gardiner (2011).

A third potential mechanism may result from binding of the GAG chains to ligands harboring growth-inhibiting or repulsive effects. There is evidence that CS GAGs interact with growth modulating ligands (Deepa et al.,2002). Candidate ligands for CS GAGs in the CNS are reviewed by Crespo et al. (2007). However, direct evidence for such interactions in vivo is scarce. One exception is the location of the soluble guidance molecule semaphorin 3A to PNNs (De Wit et al.,2005; Vo et al., unpublished results). Sema3A localizes to PNN structures and the localization depends on the CS-GAG moiety. With the use of ELISA, our lab has recently demonstrated the specific interaction of Sema3A with CS-E which is enriched in the PNNs (Deepa et al.,2006 and unpublished data). Deposition of the repulsive guidance molecule Sema3A in the PNNs could certainly explain their effects on growth inhibition and synapse dynamics. This suggests that the biological effect of the ligand–GAG interaction may change depending on the GAG type and structure. Such an example has been reported by Kantor et al. that the change of the guidance effects of sema5A from attractive to repulsive, depending on CS or HS GAG interaction (Kantor et al.,2004).


In conclusion, the ECM and PNNs are important regulators of CNS plasticity, both during development and into adulthood. There is increasing evidence suggesting that production of critical PNN components is triggered by activity such as sensory experience during the early phase of development; PNN deposition around the cell bodies and dendrites of inhibitory neurons helps to stabilize the established neuronal network and reduce the plastic changes within the CNS to future experiences. Disruption of the PNN by removal of inhibitory CSPGs, neutralization with antibodies and peptides, blocking of receptors or manipulations of PNN components, may allow activity-dependent changes to once again modify the CNS and may be helpful following CNS injury.