Perineuronal Nets and Schizophrenia: The Importance of Neuronal Coatings

Abstract

Schizophrenia is a complex brain disorder associated with deficits in synaptic connectivity. The insidious onset of this illness during late adolescence and early adulthood has been reported to be dependent on several key processes of brain development including synaptic refinement, myelination and the physiological maturation of inhibitory neural networks. Interestingly, these events coincide with the appearance of perineuronal nets (PNNs), reticular structures comprised of components of the extracellular matrix that coat a variety of cells in the mammalian brain. Until recently, the functions of the PNN had remained enigmatic, but are now considered to be important in development of the central nervous system, neuronal protection and synaptic plasticity, all elements which have been associated with schizophrenia. Here, we review the emerging evidence linking PNNs to schizophrenia. Future studies aimed at further elucidating the functions of PNNs will provide new insights into the pathophysiology of schizophrenia leading to the identification of novel therapeutic targets with the potential to restore normal synaptic integrity in the brain of patients afflicted by this illness.

Introduction

Schizophrenia is a severe and persistent mental disorder associated with a large amount of suffering, discrimination and stigma (Gerlinger et al., 2013; Phelan et al., 1998; Omori et al., 2014). The symptoms of the disorder can be divided into three broad categories (i.e. positive, negative and cognitive symptoms), which influence a patient’s thoughts, speech, perceptions, emotion and behaviors (Schultz et al., 2007). The end result is a brain disorder associated with a life-long course of intellectual, vocational, interpersonal and social impairment. Worldwide, the prevalence for schizophrenia appears to be approximately 1% (Perala et al., 2007), although variation in prevalence and incidence of the illness has been reported according to geographical region and cultural background (Kirkbride et al., 2012; McGrath et al., 2004; McGrath and Susser, 2009; Saha et al., 2005). The combined economic and social costs of this illness are significant and far reaching with schizophrenia comprising roughly 1% of the global burden of disease, a fraction that is considered moderate to high (Lora et al., 2012). Specifically, schizophrenia has been reported to make up more than 7% of the burden attributable to neuropsychiatric disorders as a whole (Whiteford et al., 2013).

While the exact mechanism underlying the pathogenesis of schizophrenia remains shrouded in mystery, the neurodevelopmental theory of schizophrenia (Murray and Lewis, 1987; Weinberger, 1987) has been the dominant paradigm for schizophrenia research over the past two decades. One interpretation of this theory posits that schizophrenia is the behavioral manifestation of a disruption in the developmental trajectory of the brain during early life that underlies the later emergence of psychosis during adulthood, stemming from a combination of small but impactful genetic and environmental factors.

Many candidate vulnerability genes for schizophrenia, including Disrupted-in-schizophrenia 1, Neuregulin 1, in addition to genes located in the cytogenetic region 22q11, have a well-documented function in regulating neurodevelopmental processes (Bradshaw and Porteous, 2012; Brandon and Sawa, 2011; Harrison and Weinberger, 2005; Karayiorgou et al., 2010; Narayan et al., 2013). Early exposure to environmental predisposing factors, such as infection (Brown, 2012; Brown and Derkits, 2010; Brown and Susser, 2002) or malnutrition before birth (Lumey et al., 2011; Xu et al., 2009), obstetric complications (Cannon et al., 2002; Clarke et al., 2006; Margari et al., 2011; Nicodemus et al., 2008; Suvisaari et al., 2013) and frequent cannabis use during adolescence (Arseneault et al., 2004; Casadio et al., 2011; Henquet et al., 2005; Semple et al., 2005) contribute to increased vulnerability for schizophrenia. Nonetheless, the precise contribution of these prenatal, perinatal and postnatal factors and the ways in which they may interact with one another and with otherwise normal developmental events to result in the expression of the schizophrenia phenotype remains unclear.

A prominent and consistent pathophysiological characteristic that has emerged in schizophrenia involves impaired synaptic connectivity (Stephan et al., 2009) (McGlashan and Hoffman, 2000; Seshadri et al., 2013), a conclusion which has been inferred from multiple imaging and postmortem brain studies (Glantz and Lewis, 1997, 2000; Harrison, 2000; Selemon and Goldman-Rakic, 1999; Whalley et al., 2012). On the basis of these findings, the current consensus is that the symptoms associated with schizophrenia may arise from a malfunctioning synaptic communication within and between brain regions, leading to a functional disruption of the circuitry that underlies many of the cognitive and perceptive functions (Uhlhaas, 2013).

It is thought that this reduced synaptic connectivity stems from disturbances of key events of brain development during late adolescence and early adulthood, including myelogenesis (Bartzokis, 2002; Bartzokis et al., 2003; Benes, 1989) and synaptic pruning (Feinberg, 1982; Keshavan et al., 1994; McGlashan and Hoffman, 2000), a process that ultimately helps sculpt the adolescent brain into its adult form. Other work supporting the neurodevelopmental nature of schizophrenia suggests that its onset is linked to a potential disruption of the physiological maturation of inhibitory neural networks consisting of fast-spiking parvalbumin-positive gamma-aminobutric acid (GABAergic) interneurons (Gonzalez-Burgos et al., 2010; Gonzalez-Burgos and Lewis, 2008; Lewis et al., 2005; Volk and Lewis, 2002, 2013), which are critical for the integrity of various higher brain functions, such as cognition and perception, via regulation of gamma band neural synchrony (Gonzalez-Burgos and Lewis, 2008; Lewis et al., 2005; Uhlhaas et al., 2009). Disturbances of parvalbumin neurons and gamma synchrony may contribute to aberrant synaptic refinement, hence triggering schizophrenia onset (Woo et al., 2010). Notably, the maturation of parvalbumin neurons and gamma is concurrent with the postnatal developmental formation of perineuronal nets (PNNs) (Mauney et al., 2013).

PNNs represent well-organized components of the extracellular matrix (ECM) that ensheath many mature central nervous system (CNS) neurons and their axons and dendritic processes, but which are excluded from sites of synaptic contact, thereby creating “holes” that give these structures their lattice-like appearance (Alpar et al., 2006; Bruckner et al., 2006; Ojima et al., 1995). Although their role is still not completely understood, PNNs have been implicated in ion homeostasis around highly active neurons (Bruckner et al., 1993; Bruckner et al., 1996a; Bruckner et al., 1996b; Hartig et al., 1999; Hartig et al., 2001; Hobohm et al., 1998) and neuroprotection (Bruckner et al., 1999; Cabungcal et al., 2013; Morawski et al., 2010; Morawski et al., 2012; Morawski et al., 2004; Schuppel et al., 2002). In addition, several studies have indicated that PNNs are important for memory and learning (Bruckner et al., 2000; Bukalo et al., 2001 Gogolla et al., 2009; Lee et al., 2012; Romberg et al., 2013; Saghatelyan et al., 2001). Moreover, PNNs appear to function in a multitude of key physiological processes during development, including the closure of the critical period (Bavelier et al., 2010; Friauf, 2000; Galtrey and Fawcett, 2007; Guimaraes et al., 1990; McRae et al., 2007; Miyata et al., 2012; Pizzorusso et al., 2002, 2006; Rauch, 2004)(Nabel and Morishita, 2013; Takesian and Hensch, 2013) as well as synaptic stability and plasticity (Dityatev and Schachner, 2003; Galtrey and Fawcett, 2007; Kwok et al., 2008; Pizzorusso et al., 2002, 2006; Wang and Fawcett, 2012)(Caroni et al., 2012), all factors that have been linked to schizophrenia pathogenesis. Of additional interest is the recent evidence linking genes encoding structural constituents of PNNs and those that regulate their formation to schizophrenia, therefore adding further support to the emerging role of these structures in the pathophysiology of this illness (Buxbaum et al., 2008; Kahler et al., 2011; Muhleisen et al., 2012; So et al., 2009; Takahashi et al., 2011a)(Pietersen et al., 2014a,b)(Ohi et al., 2013)(Ripke et al., 2013).

In this review, we articulate the importance of PNNs in maintaining synaptic integrity and how functional deficits in these structures may contribute to schizophrenia pathogenesis. We begin by summarizing the structure and organization of PNNs, before turning our focus to the unique roles of PNNs in neural and behavioural plasticity. Finally, we discuss how different mechanisms associated with PNN functions might converge and how perturbations in these functions might contribute to the pathophysiology of schizophrenia.

Discovery of PNNs

PNNs were first described by Camillo Golgi at the end of the 19th century (Bentivoglio and Mazzarello, 2010; Celio et al., 1998)(Vitellaro-Zuccarello et al., 1998)(Golgi, 1898) (Spreafico et al., 1999). However, the opinion of Ramon y Cajal, a key proponent of the neuron doctrine and an influential figure in the realm of neuroanatomy at the time, that PNNs represented a fixation artifact of metallic impregnation abruptly cut short the first period of their investigation (Bentivoglio and Mazzarello, 2010; Celio et al., 1998)(Vitellaro-Zuccarello et al., 1998). Only over the past few decades has a resurgence of interest in the PNN as a structural component of the CNS occurred thanks to the advances in the field of cytochemistry. In this respect, research efforts from pioneers in the field such as Richard and Rene Margolis have not only been able to confirm the existence of PNNs, but have also proven integral in deciphering their chemical nature, thus proving decisively that PNNs consist of condensed proteoglycans (PGs) (Margolis and Margolis, 1993, 1997).

Structural Components of PNNs

Extracellular space, of which the ECM is the principal component, comprises approximately 20% of the brain volume in the adult (Sykova et al., 1998; Sykova and Nicholson, 2008). The components of the ECM are produced intracellularly and secreted from neurons and glial cells to form a dense network of proteins and glycans that occupy the extracellular space (Franco and Muller, 2011; Frantz et al., 2010). Besides providing a structural support system for neural cells by acting as a scaffold and therefore facilitating the organization of these cells into distinct CNS regions, the ECM also represents a source of diverse molecular cues that guide cellular and neuritic growth, activity and survival (Barros et al., 2011; Franco and Muller, 2011). In particular, distinctive differences have been shown to exist between the molecular compositions of neural ECM and ECM from other tissues. Whereas the major constituents of a non-neural ECM are glycosoaminoglycan (GAG) sugars, adhesive glycoproteins (laminin and fibronectin) and fibrous proteins (collagen and elastin), the neural ECM, in contrast, is composed of low levels of fibrous proteins, high amount of GAGs (Novak and Kaye, 2000) as well as PGs, specialized glycoproteins that are negatively charged under physiological conditions (Barros et al., 2011; Galtrey and Fawcett, 2007; Margolis and Margolis, 1993; Margolis and Margolis, 1997; Ruoslahti, 1996; Yamaguchi, 2000).

The PNN constitutes a specialized assemblage of ECM molecules expressed in the CNS. The main components are hyaluronan (Figure 1), chondroitin sulfate proteoglycans (CSPGs) (i.e. aggrecan, brevican, neurocan, versican and phosphocan) (Figure 2), tenascins (Figure 3), and link proteins (Kwok et al., 2011; Lau et al., 2013; Soleman et al., 2013; Zimmermann and Dours-Zimmermann, 2008)(Kwok et al., 2012). Through a series of complex interactions, these ECM molecules form large and stable lattice-like structures that rest on the surface of somata, dendrites and axons of neurons (Figure 4). The dynamic assembly and remodeling of PNNs has been associated with the formation of synapses and connections and thereby neural plasticity (Celio et al., 1998; Kwok et al., 2011; Soleman et al., 2013; Wang and Fawcett, 2012).

Figure 1.

Chemical structure of Hyaulronan

Figure 2.

Structures of the Lectican family of chondroitin sulfate proteoglycans

Figure 3.

Domain structure of Tenascin-C and -R: Two key components of the perineuronal net

Figure 4.

Schematic of perineuronal net organization through interactions with hyaluronan, link protein, tenascins and chondroitin sulfate proteoglycans

Distribution and Development of PNNs in the CNS

A variety of histo- and immunohistochemical staining methods exist to identify PNNs, including the use of plant lecticans Vicia villosa agglutinin and Wisteria Floribunda agglutinin (WFA), which have high affinity for N-acetylglucosamine residues (Bruckner et al., 1993; Bruckner et al., 1996a; Schweizer et al., 1993), the use of colloidal iron hydroxide staining for detection of polyanionic components (Seeger et al., 1994) and monoclonal antibodies to CSPGs (with comparative studies providing evidence indicating species-specific distributions of PNNs) (Atoji et al., 1997; Bruckner et al., 1993; Celio et al., 1998; Nakagawa et al., 1986; Wintergerst et al., 1996)(Cant and Benson, 2006; Deepa et al., 2006; Hartig et al., 2001; Hilbig et al., 2007; Morawski et al., 2009; Morris and Henderson, 2000; Pantazopoulos et al., 2008; Pantazopoulos et al., 2010; Wagoner and Kulesza, 2009). While most studies of PNNs utilize WFA as a marker, PNNs exhibit a high degree of constitutive heterogeneity, and therefore it should be kept in mind that the alteration of a single marker during development or in disease states does not necessarily reflect the alteration or loss of the entire structure.

A large number of animal studies have shown that PNNs are widely distributed in the brain (Bertolotto et al., 1991; Bruckner et al., 1996b; Hockfield et al., 1990)(Dick et al., 2013; Giamanco et al., 2010; Hartig et al., 1994; Hartig et al., 1992; Nakamura et al., 2009; Yasuhara et al., 1994). Similarly, in humans, PNNs have also been found to be present in a variety of brain regions, including entorhinal cortex (Pantazopoulos et al., 2010), amygdala (Pantazopoulos et al., 2008; Pantazopoulos et al., 2010), hippocampus (Lendvai et al., 2013), motor and somatosensory cortex (Hausen et al., 1996), visual cortex (Mauney et al., 2013; Murakami et al., 1994; Seeger et al., 1996) and prefrontal cortex (Mauney et al., 2013).

Consistent with the evidence from animal studies reporting an incremental pattern of PNN expression asssociated with the postnatal maturation of the CNS (Hensch, 2005a; Miyata et al., 2012; Nabel and Morishita, 2013)(Bavelier et al., 2010; Gogolla et al., 2009; Lee and Sawatari, 2011; Nowicka et al., 2009), we have recently shown that the number of PNNs in the human prefrontal cortex also increases through the peripubertal period until late adolescence and early adulthood (Mauney et al., 2013) (see Figure 5). Formation of PNNs has been found to be activity-dependent. A series of in vitro experiments conducted on slices from mouse visual cortex showed that blockage of sodium channels inhibits the formation of PNNs (Dityatev et al., 2007). Interestingly, the same study revealed that blockage of glutamate receptors and presynaptic N- and P/Q-type calcium channels (both implicated in glutamate release) did not affect PNN growth, indicating that neuronal activity is required but that glutamatergic neurotransmission is not essential for PNN development (Dityatev et al., 2007). Other evidence stemming from the visual (Guimaraes et al., 1990; Kind et al., 1995; Lander et al., 1997; Sur et al., 1988), motor (Kalb and Hockfield, 1988, 1990a, b), and somatosensory system (McRae et al., 2007) in animals, as well as in the pallial (cortical) song nuclei (Balmer et al., 2009) in songbirds suggest that PNN expression does not only revolve around neuronal activity, but rather its expression is dependent on neuronal activity during a maturationally based ‘critical period’, when circuits become particularly sensitive to environmental/activity dependent manipulation (Hensch, 2005b; McRae and Porter, 2012; McRae et al., 2007; Wiesel and Hubel, 1963)(Nabel and Morishita, 2013) (Takesian and Hensch, 2013). The exact mechanisms that govern the ability of neuronal activity to regulate the expression of the PNN remains an open question. Likewise, it is also unknown as to how the adult PNN is able to withstand changes in neuronal activity, making them adequately suited for its role in stabilizing synapses following the advent of neuronal circuitry maturity (Zaremba et al., 1989).

Figure 5. Postnatal development of perineuronal nets in the human prefrontal cortex.

A. Photomicrographs demonstrating the increase in perineuronal nets (PNNs) in the prefrontal cortex during postnatal development. B. Linear regression analysis indicates statistically significant effect of age on PNN density in the entire prefrontal cortex (R²=0.45, p=0.0017) and in layer 3 (R²=0.49, p=0.0008), suggesting that the density of PNNs in the prefrontal cortex undergoes a prolonged course of progressive increase during postnatal development through adolescence and early adulthood. However, the nonlinear hyperbolic regression models appear to be a better fit of the data (R²=0.71 and 0.76 for the entire prefrontal cortex and layer 3, respectively); these models suggest that PNN density increases during postnatal development with the most pronounced changes occurring around the peri-pubertal period. These findings were derived from postmortem human brains from 19 healthy control subjects obtained from the National Institute of Child and Human Development Brain and Tissue Bank at the University of Maryland in Baltimore, MD. Reproduced with permission of Springer-Verlag, Heidelberg.

The Role of PNNs in Regulating Plasticity

In recent years, accumulating data suggest that the brain can be surprisingly plastic following synaptic stabilization in the mature organism, with intriguing mechanisms of neurogenesis and synaptogenesis providing novel substrates for brain adaption (Kelsch et al., 2010; Ming and Song, 2011; Song et al., 2005; Bavelier et al., 2010). This is not a new idea per se, and the reader is referred to many more detailed reviews that dissect mechanisms underlying the role of PNN function in CNS plasticity (Dityatev and Schachner, 2003; Galtrey and Fawcett, 2007; Kwok et al., 2011; Wang and Fawcett, 2012). Here we provide a summary of the experimental evidence implicating PNNs in the regulation of synaptic plasticity during critical period development.

The visual system has been the preeminent model of experience-dependent critical period plasticity (Sur et al., 2013). Monocular deprivation in young animals during the critical period leads to a shift in ocular dominance favoring the non-deprived eye; however, in the adult, due to the less plastic nature of the brain, this shift does not occur. To address the role of PNNs in ocular dominance plasticity, Pizzorusso and colleagues used chondroitinase ABC (ChABC) to enzymatically degrade these structures in adult rats (Pizzorusso et al., 2002). Remarkably, this treatment resulted in the reinstatement of ocular dominance plasticity in monocular deprived adult animals (Pizzorusso et al., 2002). In other words, by removing the PNN, synaptic plasticity normally seen only during the critical period was restored (Pizzorusso et al., 2002). Furthermore, when ChABC was applied in adult animals that had undergone monocular deprivation in combination with reverse lid suturing (the previously deprived eye is opened and the non-deprived eye is sutured) during critical period, there was a complete recovery of visual acuity and dendritic spine density (Pizzorusso et al., 2002). Together these findings provide strong support for the notion that PNN formation during development regulates the termination of the critical period in the visual cortex. The mechanism by which PNNs may regulate critical period plasticity in the visual system is not well understood. However, recent evidence has shown the PNN to play a role in the “capture” of the homeodeomain transcription factor orthodenticle homeobox 2 (Otx2) from thalamic inputs to visual cortex (Beurdeley et al., 2012). Notably, this homeoprotein signals the maturation of the parvalbumin-containing GABAergic interneurons, which contributes to the opening and possibly also the closure of the critical period (Beurdeley et al., 2012; Miyata et al., 2012; Spatazza et al., 2013; Sugiyama et al., 2008).

The somatosensory system in rodents has also been widely used as a model for studying the formation and plasticity of specific neuronal connections in the brain. By manipulating this system in mice, McRae and co-workers showed that the expression of PNNs in the mouse barrel cortex during development was dependent on appropriate sensory stimulation (McRae et al., 2007). Specifically, sensory deprivation by whisker trimming within the first 30 days of birth was found to decrease the number of aggrecan-positive PNNs around parvalbumin-expressing cells in layer IV of the mouse barrel cortex and prolong the critical period (McRae et al., 2007). Sensory deprivation in adult mice, however, did not alter aggrecan expression, indicating that once established, maintenance of PNNs does not require normal sensory input (McRae et al., 2007). Similarly, sensory deprivation of facial vibrissae in rat barrel cortex has also been found to result in reduced numbers of WFA-labeled PNNs (Nakamura et al., 2009). These observations are reminiscent of the cat visual system, where aggrecan expression decreases after sensory deprivation during development but not in adulthood (Sur et al., 1988).

Converging lines of evidence indicate that the amygdala plays a crucial role in the development and expression of conditioned fear and extinction (Maren et al., 2013; Pare and Duvarci, 2012). Against this background, it was recently reported that PNNs in the amygdala may serve a key role in the developmental regulation of fear extinction (Gogolla et al., 2009). Because the maturation of PNNs in the mouse amygdala was shown to coincide with the end of the developmental period during which extinction induces erasure of fear memories, it was suggested that PNNs prevent extinction from erasing fear. Interestingly, enzymatic degradation of PNNs in the basolateral nucleus of the amygdala of adult mice with ChABC indeed enabled extinction to erase conditioned fear, but only if PNNs were degraded before fear conditioning (Gogolla et al., 2009). This effect was, however, specific for fear memories acquired in the absence of PNNs, as the elimination of PNNs in adult mice that acquired fear with PNNs intact had no effect on extinction. These findings indicate that PNNs are actively involved in protecting the acquisition of fear memories from extinction-induced erasure, allowing extinction memories to coexist with previously acquired fear memories (Gogolla et al., 2009; Quirk et al., 2010).

Recently, a depletion of PNNs in animals following localized digestion with ChABC was shown to result in enhanced object recognition, in addition to a facilitation of long-term depression in the perirhinal cortex, thought to be the major synaptic mechanism underlying object recognition memory (Romberg et al., 2013). Interestingly, an identical prolongation of memory was observed in Ctrl1 (cartilage link protein) knockout mice, in which PNN formation is disrupted (Carulli et al., 2010; Vo et al., 2013). This study also revealed that ChABC induced enhancement of object recognition was reversible. Notably, over the course of 8 weeks, memory gradually returned to control levels, suggesting that reoccurring PNNs gradually restore control plasticity levels (Romberg et al., 2013).

Other evidence implicating PNNs in normal behavior stems from the colocalization of these structures in the striatum (Lee et al., 2012). Indeed, bilateral digestion of striatal PNNs was reported to increase both the width and variability of hind limb gait. Intriguingly, this also resulted in an improvement in the acquisition rate of the Morris water maze, a behavioural test used for assessing spatial and related forms of learning and memory (Morris, 1984). These findings suggest that PNNs are associated with specific elements of striatal circuits and play a key role in regulating the function of this important structure (Lee et al., 2012).

Finally, the role of PNNs in synaptic and behavioral plasiticity appears to extend beyond mammalian species. Evidence from the songbird (zebra finch) suggests the functional role of PNNs in vocal plasticity. A study by Balmer and colleagues found the percentage of both total and parvalbumin-positive neurons with PNNs increased with development (Balmer et al., 2009). In the high vocal center, a song area important for sensorimotor integration (During et al., 2013; Wild, 2004), the percentage of parvalbumin neurons with PNNs correlated with song maturity. Notably, shifting the vocal critical period with tutor song deprivation decreased the percentage of neurons that were parvalbumin positive and the relative staining intensity of both parvalbumin and PNNs (Balmer et al., 2009). Because developmental song learning shares key characteristics with sensory critical periods, the songbird system represents an efficient model in which to examine the mechanisms of critical period regulation.

Linking PNNs to Schizophrenia Pathophysiology

Because of the critical function of PNNs in synaptic plasticity, disruption of these structures may contribute to the neuronal circuitry dysfunction associated with neurological and psychiatric diseases (Lau et al., 2013). In this context, a number of studies have shown alterations in PNNs and their constituents to be linked to various brain diseases, such as Alzheimer’s disease (Morawski et al., 2010, 2012; Kobayashi et al., 1989; Baig et al., 2005), epilepsy (McRae and Porter, 2012), multiple sclerosis (Gray et al., 2008), prion disease (Guentchev et al., 1999, Franklin et al., 2008) and stroke (Hobohm et al., 2005; Gherardini et al., 2013; Soleman et al., 2012). Recent evidence suggests that PNNs may also be involved in the pathophysiology of schizophrenia. Specifically, a number of human post-mortem brain studies have revealed a disease-specific reduction in the density of PNNs as well as altered expression of genes that regulate PNNs and ECM in key brain structures associated with schizophrenia, including the amygdala, olfactory epithelium, entorhinal cortex, superior temporal cortex and prefrontal cortex (Mauney et al., 2013; Pantazopoulos et al., 2013; Pantazopoulos et al., 2010; Pietersen et al., 2014a, b) (See Figure 6). In this section, we discuss these findings in the context of known and postulated PNN functions and speculate elements of these functions that might play a role in the synaptic pathology associated with this illness.

Figure 6. Densities of perineuronal nets in the prefrontal cortex in subjects with schizophrenia.

A. Representative photomicrographs showing the distribution of perineuronal nets (PNNs) in the prefrontal cortex in a schizophrenia (right) and a normal control (left) subjects. Scale bar=100μm. B. Photomicrograph showing a WFA-labeled PNN. Scale bar=20μm. C. WFA-labeled PNNs are significantly decreased in layers 3 (70%) and 5 (76%) in subjects with schizophrenia (SZ; N=16). Bar graphs represent the mean and upper 95% confidence interval by cortical layer. Layer 1 is not shown because no PNNs were found in that layer. There are no significant differences in PNN densities between subjects with bipolar disorder (N=15) and normal control (N=16) subjects. p value, F ratio: *(0.016, 6.49); **(0.028, 5.36); ***(0.042, 4.51). These findings were derived from postmortem human brains obtained from the Harvard Brain Tissue Resource Center in Belmont, MA. Reproduced with permission of Springer-Verlag, Heidelberg.

Ion Homeostatsis

Based on their polyanionic character, PNNs provide highly charged structures in the direct microenvironment of neurons that might be involved in local ion homeostasis (Karetko and Skangiel-Kramska, 2009; Morawski et al., 2004). They can thus potentially act as a buffering system for physiologically relevant ions such as calcium, potassium, and sodium around highly active types of neurons (Bruckner et al., 1993; Bruckner et al., 1996b; Hartig et al., 1999), although direct evidence to this effect is still lacking. Even though PNNs have been shown to ensheath a variety of cells in the CNS, they are most commonly associated with the highly metabolically active fast-spiking parvalbumin interneurons (Hartig et al., 1999; Morris and Henderson, 2000; Ojima et al., 1995; Wegner et al., 2003)(Liu et al., 2013). These neurons express specific potassium channel subunits, such as Kv3.1b, which produce rapidly repolarizing action potentials (Lenz et al., 1994)(Du et al., 1996; Sekirnjak et al., 1997), thus contributing to the fast-spiking phenotype. The firing properties of parvalbumin interneurons participate in establishing the normal gamma oscillations and synchrony of cortical neuronal populations, thought to be a physiological correlate of higher-order information processing in brain (Gray et al., 1992; Singer, 1990; Buzsaki, 2006; Uhlhaas et al., 2008; Uhlhaas and Singer, 2006). In this regard, schizophrenia is associated with deficits in parvalbumin neurons (Bitanihirwe et al., 2009; Hashimoto et al., 2003), abnormalities in cortical gamma synchrony (Basar-Eroglu et al., 2007; Cho et al., 2006; Farzan et al., 2010; Grutzner et al., 2013; Kirihara et al., 2012; Spencer et al., 2008; Sun et al., 2011; Uhlhaas and Singer, 2010; Woo et al., 2010) and in information processing (Braff, 1993; Frith, 1979; Kumari et al., 1999; Perry et al., 2002; Rissling et al., 2012), particularly with dysfunction in working memory and executive function (Goldman-Rakic, 1994; Gonzalez-Burgos et al., 2010; Gonzalez-Burgos and Lewis, 2008; Lewis et al., 2005). Considering that Kv3.1b has a direct effect on the firing of the parvalbumin interneurons (Gan and Kaczmarek, 1998; Lenz et al., 1994; Perney et al., 1992; Weiser et al., 1995), one may speculate that deficits in PNNs may affect local ion homeostasis due to excess potassium extruded by the cell during activity (Bruckner et al., 1993), subsequently resulting in disrupted firing properties of neurons (i.e. neuronal hyperexcitability). In this respect, a variety of evidence exists indicating subtle cortical hyperexcitability in patients with schizophrenia (Daskalakis et al., 2007; Eichammer et al., 2004; Hoffman et al., 2002; Hoffmann and Cavus, 2002; Spencer et al., 2009; Lakatos et al., 2013). In combination with anomalies in the expression of the Kv3.1b (Yanagi et al., 2013) and other potassium channels (Georgiev et al., 2014; Georgiev et al., 2012), distorted ion homeostasis resulting from PNN deficit could be important in mediating alterations in brain function associated with schizophrenia.

The nodes of Ranvier represent specialized gaps along the length of the axon stemming from an absence of myelin, produced by oligodendrocytes as part of their supportive role to neurons (Bradl and Lassmann, 2010; Wilkins et al., 2003). The myelin sheath reduces current flow across the axonal membrane by providing a layer of high electrical resistance and low capacitance, thereby enabling a fast, saltatory movement of nerve impulses from ‘node to node’ (Poliak and Peles, 2003). In this context, multiple lines of evidence in schizophrenia, from electrophysiological, genetic analyses, brain imaging and studies in postmortem brains have implicated oligodendrocyte and myelin dysfunction in this disease (Benes, 2000; Hoistad et al., 2009; Takahashi et al., 2011b). Particularly noteworthy are the molecular studies showing a reduction in expression level of multiple genes associated with the integrity of the node of Ranvier (Martins-de-Souza, 2010; Martins-de-Souza et al., 2009; Roussos et al., 2012; Susuki, 2013). Notably, the perinodal-matrix or peridonal net, a PNN-like structure (Dours-Zimmermann et al., 2009; Zimmermann and Dours-Zimmermann, 2008), is commonly associated with the node of Ranvier. This structure functions to create an ion diffusion barrier by acting as a cation buffer at the nodes of Ranvier (Bekku et al., 2010). Given that transgenic animals lacking the CSPG brevican have been reported to exhibit disruptions in the perinodal-net in addition to deficits in nerve conduction (Bekku et al., 2010), it can be conjectured that a disruption in the expression of perinodal-net components, such as brevican, may contribute to the myelin and electrophysiological deficits observed in schizophrenia.

Neuronal Protection

The potentially neuroprotective function of PNNs might be understood against the background of its molecular interactions with elements associated with neurotoxicty, such as metal ions (Karetko and Skangiel-Kramska, 2009). Because of the polyanionic character of PNNs, it has been suggested that they may provide a system for scavenging and binding redox-active ions, such as iron, copper and zinc, and thus reducing the local oxidative potential in the neuronal microenvironment (Morawski et al., 2004)(Karetko and Skangiel-Kramska, 2009). This may provide some neuroprotection to PNN-associated neurons against oxidative stress. Oxidative stress is a pathophysiological process resulting from an impairment in cellular antioxidant defense mechanisms to counterbalance and control endogenous reactive oxygen species and reactive nitrogen species generated from normal oxidative metabolism or from pro-oxidant environmental exposures (Kohen and Nyska, 2002; Roberts et al., 2009, 2010). Notably, this process has been linked to the pathophysiology of various neurological disorders including schizophrenia (Bitanihirwe and Woo, 2011; Do et al., 2009; Yao and Reddy, 2011).

A recent study demonstrated the importance of PNN integrity in reducing the pathophysiologcal sequalae of oxidative stress (Cabungcal et al., 2013). This study used a mouse model in which disrupted expression of the modifier subunit of glutamate cysteine ligase, the rate-limiting enzyme of glutatione synthesis, elicits elevated oxidative stress (Yang et al., 2002). These animals were challenged with a 6 month chronic oxidative insult using a specific dopamine reuptake inhibitor (GBR12909). The authors found that animals with well formed PNNs (viz. high levels of WFA-labeling) were protected against chronic redox dysregulation. Furthermore, PNN deficit within these mice preferentially affected parvalbumin-containing inhibitory interneurons with a concomitant reduction in neural oscillations in the beta and gamma frequency bands (Cabungcal et al., 2013). This study reported that enzymatic degradation of PNNs with ChABC rendered mature parvalbumin cells and fast rhythmic neuronal synchrony more susceptible to oxidative stress. Interestingly, the frequency of gamma band oscillations may actually be increased in mice with a mild PNN deficit (ChABC treatment alone) while the frequency of gamma band oscillations is decreased (compared to controls) once more oxidative stress is placed on the system. In this regard, one potential interpretation of the human PNN postmortem findings could be that the decrease in PNNs observed in patients with schizophrenia may represent a compensation to elevate gamma band oscillation and neuronal synchrony. Consisitent with the notion that the expression of PNNs is accompanied by the maturation of the CNS, parvalbumin-containing cells in juvenile mice showed higher susceptibility to oxidative stress compared with adult mice (Cabungcal et al., 2013). Together these findings highlight the possible role of PNNs in protecting the fast-spiking parvalbumin cells against oxidative stress, which in turn prevents the electrophysiological deficits associated with schizophrenia.

The dopamine hypothesis of schizophrenia stems from the body of work suggesting that the major antipsychotic drugs act by blocking dopamine D2 receptors and that dopamine-releasing drugs worsen schizophrenia symptoms (Snyder, 1976). In this context, rats treated with intra-ventral hippopcampal ChABC to remove PNNs not only exhibited an increase in dopamine neuron population activity, as shown by a significantly greater firing rate of putative pyramidal neurons throughout the ventral hippocampus, but were also hyper responsive to the locomotor stimulating effects of amphetamine (Shah and Lodge, 2013). In a complimentary set of experiments, the offspring of rat dams exposed to the DNA-alkylating agent methylazoxymethanol acetate had significantly lower levels of the PNN proteins brevican and phosphacan throughout the ventral hippocampus (Shah and Lodge, 2013). Together these data suggest that a loss of hippocampal PNNs is sufficient, in and of itself, to augment hippocampal activity and induce the dopamine system hyperfunction purported to underlie the positive symptoms of schizophrenia (Howes and Kapur, 2009).

Genetic Manipulation of Murine Phosphacan and its Relevance to Schizophrenia

The spatiotemporal expression pattern of phosphacan within the CNS suggests potential roles for this protein in various developmental processes, including cell migration (Maeda et al., 1996), neurite growth (Garwood et al., 2003), synaptogenesis (Haunso et al., 1999), synaptic function (Murai et al., 2002), and myelination (Harroch et al., 2000)(Garwood et al., 2001), all processes that have been postulated to be affected in schizophrenia. Nonetheless, the role of phosphacan in schizophrenia is fragmentary. On the basis of a genetic case-control study that revealed association between the PTPRZ1 gene (which encodes RPTPβ also known as phosphacan) and schizophrenia (Buxbaum et al., 2008), postmortem brain analysis and transgenic animal studies were subsequently performed to gain a better insight into the biological relevance of this gene to schizophrenia. Expression of RPTPβ at the mRNA level was found to be significantly higher in the dorsolateral prefrontal cortex of patients with schizophrenia (Takahashi et al., 2011a). Given the PNN deficit observed in the prefrontal cortex of patients with schizophrenia (Mauney et al., 2013), increased RPTPβ expression could represent a compensatory mechanism to counterbalance the aberrant PNN formation in the brains of patients with this disease. The limitation, however, is that because the finding of increased expression of RPTPβ was generated from homogenized brain tissue, it is not possible to determine any cell type-specific changes. Notably, the same study found that transgenic animals overexpressing RPTPβ exhibited altered glutamatergic, GABAergic and dopaminergic activity, as well as delayed oligodendrocyte development, which are perhaps more relevant in the schizophrenia context (Takahashi et al., 2011a). Finally, these animals also demonstrated schizophrenia-like behavioral deficits, including reduced sensory motor gating, hyperactivity and working memory impairment, indicating that enhanced RPTPβ signaling can contribute to schizophrenia phenotypes (Takahashi et al., 2011a).

Development and Synaptic plasticity

Given that PNNs can modultate neural circuitry functions stemming from developmental plasticity and closure of the critical period (Friauf, 2000; Guimaraes et al., 1990; Pizzorusso et al., 2002, 2006)(Galtrey and Fawcett, 2007)(Rauch, 2004)(McRae et al., 2007)(Takesian and Hensch, 2013) to maintenance of synaptic network stability in adulthood (Dityatev and Schachner, 2003; Kwok et al., 2008)(Pizzorusso et al., 2002, 2006)(Galtrey and Fawcett, 2007)(Caroni et al., 2012), their disturbances can have widespread consequences, depending in part on when and where these disturbances occur.

In light of the fact that PNNs are known to play a role in regulating developmental synaptic pruning in the cerebral cortex and that synaptic pruning deficit has long been speculated to contribute to the onset of schizophrenia (McGlashan and Hoffman, 2000; Feinberg, 1982; Keshavan et al., 1994), we speculate that a deficit in the formation of PNNs that ensheath parvalbumin and pyramidal neurons and their processes could compromise the experience-dependent consolidation of synaptic connectivities and prolonging the termination of synaptic pruning. In this regard, it is noteworthy that a series of elegant experiments by Orlando and colleagues recently found that removal of PNNs in the hippocampus using ChABC resulted in altered spine dynamics via a restriction of β1-integrin activation and signaling at synaptic sites (Orlando et al., 2012). This is an interesting observation in light of the finding of reduction in spine number on layer 3 pyramidal cells in schizophrenia (Glantz and Lewis, 2000; Sweet et al., 2009).

Yet another element that may be linked to schizophrenia pathophysiology is the biophysics of the neural-ECM. Because diffusion in the extracellular space is dependent on the structure and chemical properties of the ECM (Sykova, 2005), it is conceivable that PNN abnormalities in the adult CNS may alter the efficacy of synapses and transmitter release associated with schizophrenia (Earls and Zakharenko, 2013). Such changes could affect the efficacy of signal transmission at synapses by altering neuronal synchronization and neuron-glia communication but also by affecting synaptic and extrasynaptic volume transmission (Fuxe et al., 2013; Sykova, 2005), a notable example being dopamine (Fuxe et al., 2013).

Finally, Nogo receptor (NgR), which binds myelin-associated proteins that inhibit axonal growth, such as the Nogo proteins, is expressed in PNN-encapsulated neurons (Ye and Miao, 2013) and also binds CSPGs (Dickendesher et al., 2012). Hence, together with PNNs, myelin formation may provide an additional mechanism that progressively restrcits synaptic plasticity and thereby stabilize cortical circuitry in the mature brain. Deficit in myelination in schizophrenia may therefore further compromise the stability and integrity of synaptic connectivities. Interestingly, Nogo and NgR have previously been implicated in the pathophysiology schizophrenia (Willi and Schwab, 2013)(Willi et al., 2010)(Tews et al., 2013)(Budel et al., 2008).

Possible Neurobiological Basis of PNN Deficit in Schizophrenia

The mechanisms underlying PNN deficit, which may be a consequence of PNN degradation and/or failure in PNN formation, in schizophrenia are unclear. It is possible that the degradation of the PNN is a consequence of an alteration in proteases involved in regulating the dynamic nature of the ECM. Two families of endogenous, extracellular metalloproteinases cleave ECM components: matrix metalloproteinase (MMPs) and a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTSs). Because cleavage of ECM components is part of their normal turnover process, a dysregulation in this process may be a prerequisite for neurological disease. In this context, several observations suggest that the MMPs may be key mediators of PNN degradation. Two members of this large group of ECM-degrading enzymes have been shown to have direct CSPG-degrading properties in the brain: MMP-2 and MMP-24 (Wang et al., 1999; Zuo et al., 1998), and other evidence (Muir et al., 2002) has shown that MMP-2 and MMP-3 can degrade most of the components of the PNN. In addition, degradation of brevican by ADAMTS has been demonstrated (Mayer et al., 2005). Moreover, ADAMTS-mediated brevican degradation colocalized with areas of synaptic loss in a kainic acid model of acute neuronal toxicity (Yuan et al., 2002). Interestingly, a recent genome-wide association study has identified MMP16 as a schizophrenia risk gene (Ripke et al., 2013). We have also recently found in laser-captured pyramidal neurons from layer 3 of the superior temporal gyrus of patients with schizophrenia to exhibit alterations in genes that encode both MMPs and ADAMTSs (Pietersen et al., 2014a, b), including MMP16, suggesting that a dysregulation in the remodeling of the ECM may represent a genuine component underlying the pathophysiology of the disease (Pietersen et al., 2014a, b) (see Table 1).

Table 1.

Differentially expressed genes associated with extracellular matrix in pyramidal neurons in schizophrenia.

Gene title	Gene symbol	Direction of change
aggrecan	ACAN	Down
ADAM metallopeptidase with thrombospondin type 1 motif, 1	ADAMTS1	Up
ADAM metallopeptidase with thrombospondin type 1 motif, 6	ADAMTS6	Up
hyaluronan and proteoglycan link protein 1	HAPLN1	Down
leucine proline-enriched proteoglycan (leprecan) 1	LEPRE1	Down
lumican	LUM	Down
matrix metallopeptidase 16 (membrane-inserted)	MMP16	Down
matrix metallopeptidase 25	MMP25	Down
matrix metallopeptidase 24 (membrane-inserted)	MMP24	Up
sperm adhesion molecule 1 (PH-20 hyaluronidase, zona pellucida binding)	SPAM1	Up
sparc/osteonectin, cwcv and kazal-like domains proteoglycan (testican) 3	SPOCK3	Up
spondin 1, extracellular matrix protein	SPON1	Up
versican	VCAN	Down

An alternative explanation is that the degradation of the PNN is a consequence of microglial activation. For example, degradation of the PNN is a feature of multiple sclerosis (Gray et al., 2008), a neurological condition that shares a similar clinical course, age of onset and immunological profile to patients with schizophrenia (Stevens, 1988). The degradation of the PNN in multiple sclerosis has been attributed to the production of MMPs by cytokines or activated microglia (Gray et al., 2008). In this context, the neuropathology of schizophrenia has been reported to be closely associated with cytokines and microglial activation (Bayer et al., 1999; Brown, 2011; Juckel et al., 2011; Meyer et al., 2009; Monji et al., 2009; Najjar et al., 2013; Radewicz et al., 2000; Vuillermot et al., 2010). Although speculative, these observations suggest that immune dysfunction may be associated with the PNN deficits observed in schizophrenia. Experimental evidence confirming such a link would be necessary to validate this conjecture. In this regard, prenatal exposure to the immunostimulant polyinosinic:polycytidilic acid has been shown to represent a robust experimental model with a well-characterized evolution of pathologic and behavioural changes relevant to schizophrenia (Bitanihirwe et al., 2010; Giovanoli et al., 2013; Nyffeler et al., 2006; Richetto et al., 2013; Vuillermot et al., 2010)(Meyer, 2013). The use of such a tool would prove effective in answering the question as to whether an association exists between immune system activation and loss of PNNs.

Conclusions

As the underlying pathology of schizophrenia continues to be unravelled, novel and exciting pieces to the complex puzzle continue to emerge. One such piece of the schizophrenia puzzle involves PNNs, whose location directly on the surface of neurons places them in a unique position to not only modulate neuronal physiology but to also exert a strong influence on synaptic formation, plasticity and stabilization in the brain (Kwok et al., 2008; Kwok et al., 2011; McRae and Porter, 2012; Wang and Fawcett, 2012). These lattice-like structures represent an integral part of the synapse and appear to be specialized to monitor and adapt according to the dynamic changes in the extracellular environment. Changes in the extracellular milieu under pathological conditions may therefore affect and disrupt the structural integrity of PNNs, triggering a ‘neurometabolic’ cascade of events culminating in neuronal damage and synaptic dysfunction associated with schizophrenia. Therefore, the normalization of PNN structural integrity may rescue the neurodevelopmental alterations affecting the physiology of neural circuitry in schizophrenia. In this context, a more complete understanding of PNN function in normal conditions and how their functions are altered in schizophrenia will be of importance in terms of providing new insight into the neurobiological underpinnings of this major public health problem. This information will likely lead to an emergence of novel targets for therapeutic intervention aimed at reversing underlying abnormalities in neural plasticity and in the process improve clinical outcomes.

Highlights.

• Deficits in PNNs impair key brain maturational processes linked to schizophrenia

• Constituent elements of the PNN are dysregulated in patients with schizophrenia

• Poor structural integrity of PNNs may underlie the neural deficits in schizophrenia

• Restoring the integrity of PNNs may help rescue neurodevelopmental abnormalities