Loss of Interneuron-Derived Collagen XIX Leads to a Reduction in Perineuronal Nets in the Mammalian Telencephalon

Perineuronal nets (PNNs) are lattice-like supramolecular assemblies of extracellular glycoproteins that surround subsets of neuronal cell bodies in the mammalian telencephalon. PNNs emerge at the end of the critical period of brain development, limit neuronal plasticity in the adult brain, and are lost in a variety of complex brain disorders diseases, including schizophrenia. The link between PNNs and schizophrenia led us to question whether neuronally expressed extracellular matrix (ECM) molecules associated with schizophrenia contribute to the assembly of these specialized supramolecular ECM assemblies. We focused on collagen XIX—a minor, nonfibrillar collagen expressed by subsets of telencephalic interneurons. Genetic alterations in the region encoding collagen XIX have been associated with familial schizophrenia, and loss of this collagen in mice results in altered inhibitory synapses, seizures, and the acquisition of schizophrenia-related behaviors. Here, we demonstrate that loss of collagen XIX also results in a reduction of telencephalic PNNs. Loss of PNNs was accompanied with reduced levels of aggrecan (Acan), a major component of PNNs. Despite reduced levels of PNN constituents in collagen XIX-deficient mice (col19a1−/−), we failed to detect reduced expression of genes encoding these ECM molecules. Instead, we discovered a widespread upregulation of extracellular proteases capable of cleaving Acan and other PNN constituents in col19a1−/− brains. Taken together, these results suggest a mechanism by which the loss of collagen XIX speeds PNN degradation and they identify a novel mechanism by which the loss of collagen XIX may contribute to complex brain disorders.


Introduction
Extracellular matrix (ECM) is a noncellular component of all tissues, composed of a dense mixture of proteins and polysaccharides. Outside of the nervous system, major constituents of ECM include laminins, fibrillar and nonfribrillar (i.e., unconventional) collagens, fibronectins, nidogens, tenascins, and proteoglycans (i.e., proteins with one or more covalently linked glycosaminoglycan side chains) (LeBleu et al., 2007;Frantz et al., 2010). Interactions between these ECM constituents lead to their assembly into supramolecular structures, which include basal laminas, basement membranes, filaments, and fibrils. These supramolecular ECM structures provide structural support for adjacent cells, act as reservoirs for growth factors, morphogens, and extracellular proteases, and impact nearly all biological functions outside of the brain.
Despite our vast knowledge of ECM assemblies in most tissues, our understanding of their role in brain development and function remains limited. This is surprising given that a significant portion of brain parenchyma comprises extracellular space. In the juvenile brain, when neural circuits are still being assembled, extracellular space accounts for $40% of brain parenchyma (Nicholson and Sykova, 1998). Although there is a dramatic reduction in the volume of extracellular space as the mammalian brain matures, a significant fraction of adult brain parenchyma remains occupied by extracellular space ($20%; Nicholson and Sykova, 1998;Sykova and Nicholson, 2008;Korogod et al., 2015;Pallotto et al., 2015). Despite this, there has been little attention paid to the presence and roles of supramolecular ECM assemblies in the mammalian brain. This likely stems from a general lack of fibrillar assemblies of ECM constituents in the healthy brain and a limited presence of basal laminas or basement membranes. Basal lamina-like structures in the brain are limited to vascular structures and to ''fractones,'' specialized ECM assemblies associated with the subventricular zone (Mercier et al., 2002;Mercier, 2016). However, it is becoming increasing clear that many ECM constituents are in fact generated by neurons and glia in the developing, adult, and injured brain and these constituents play essential roles in neuronal migration, neural circuit formation, and synaptic plasticity (Dityatev et al., 2010;Dityatev and Rusakov, 2011;Frischknecht and Gundelfinger, 2012;Risher and Eroglu, 2012;Heikkinen et al., 2014;Smith et al., 2015). But, at present, we lack a clear understanding of how these factors are assembled in the extracellular space surrounding neurons and glia in the brain. The one exception to this is the perineuronal net (PNN)-a specialized, lattice-like supramolecular assembly of ECM constituents that surrounds the cell bodies and proximal neurites of select classes of interneurons (Celio and Blumcke, 1994;Celio et al., 1998;Yamaguchi, 2000). PNNs emerge late in postnatal brain development, coinciding with closing of the critical period of brain development. For this reason it has been hypothesized that these specialized matrices limit the neuroplastic potential of adult brains (Rhodes and Fawcett, 2004;Faissner et al., 2010;Wang and Fawcett, 2012). PNNs have garnered attention not only for their potential role in developmental plasticity but also because they appear disrupted in a number of psychiatric and complex brain disorders.
Interestingly, the molecular composition of PNNs differs from supramolecular assemblies of ECM constituents outside the brain. PNNs are composed primarily of lecticans (a family of chondroitin sulfate proteoglycans [CSPGs]), tenascins, hyaluranonin and proteoglycan binding link proteins (HAPLNs), and hyaluran (Zimmermann and Dours-Zimmermann, 2008;Frischknecht and Seidenbecher, 2012;Morawski et al., 2012). More traditional ECM constituents, such as laminins, nidogens, fibronectin, and fibrillar and nonfibrillar collagens, are either absent from PNNs or their roles remain unclear. However, we have recently become interested in the link between collagen XIX and PNNs. Collagen XIX is a nonfibrillar, unconventional collagen generated by telencephalic interneurons in the postnatal rodent brain (Sumiyoshi et al., 1997;Su et al., 2010;Su et al., 2016). Collagen XIX-expressing neurons are present in regions densely packed with PNNs. And, just as PNNs have been linked to schizophrenia and complex brain disorders (Berretta, 2012;Berretta et al., 2015;Bitanihirwe et al., 2016), genetic alterations in the region encoding collagen XIX have been associated with schizophrenia (Liao et al., 2012). Furthermore, targeted deletion of collagen XIX in mice leads to the acquisition of schizophrenia-related behaviors and seizures (Su et al., 2016). Based upon these similarities, we tested whether collagen XIX contributes to the assembly or maintenance of PNNs by assessing PNN morphology and distribution in targeted mouse mutants lacking this collagen (col19a1 À/À ). Our data revealed a significant reduction in the number of PNNs in regions of mouse brain that normally generate collagen XIX. Loss of PNNs was accompanied by a widespread reduction in protein levels of lecticans in these regions of col19a1 À/À mutant brains. However, we failed to detect reduced levels of lectican or PNN constituent mRNA in mutants, suggesting that the reduced number of PNNs in collagen XIXdeficient brains did not result from diminished lectican transcription. Finally, we tested whether the loss of this unconventional neuronally expressed collagen altered the expression of extracellular proteases that degrade lecticans and PNNs. To our surprise, we discovered a widespread upregulation in the transcription of matrix metalloproteinases (MMPs) and a distintegrin and metalloproteinase with thrombospondin motif proteases (ADAMTSs) in these regions of col19a1 À/À mutant brains.

Immunohistochemistry
Fluorescent immunohistochemistry (IHC) was performed on 16 mm cryosectioned paraformaldehyde (PFA)-fixed brain tissue as described previously (Fox et al., 2007;Su et al., 2012). Briefly, tissue slides were allowed to air dry for 15 min before being incubated with blocking buffer (2.5% normal goat serum, 2.5% bovine serum albumin, and 0.1% Triton X-100 in PBS) for 30 min. Primary antibodies were diluted in blocking buffer and incubated on tissue sections for overnight at 4 C. On the following day, tissue slides were washed in PBS and secondary antibodies diluted 1:1000 in blocking buffer were applied to slides for 1 hr at room temperature. After thoroughly washing in PBS, tissue slides were coverslipped with VectaShield (Vector Laboratories, Burlingame, CA, USA). Images were acquired on a Zeiss Examiner Z1 LSM 710 confocal microscope or a Zeiss LSM 700 confocal microscope (Oberkochen, Germany). When comparing between genotypes, images were acquired with identical parameters and analyzed blind. A minimum of three animals (per genotype and per age) were compared in all IHC experiments. IHC images were analyzed and quantified as previously described (Su et al., 2010;Levy et al., 2015;Su et al., 2016). For PNN counts in hippocampus, all images were obtained with identical parameters, and to ensure similar regions were analyzed, we only analyzed coronal sections of hippocampus that also contained central regions of visual thalamus-which includes both dorsal and ventral lateral geniculate nuclei (dLGN and vLGN, respectively). For quantification purposes, all WFA and Acan-labeled PNNs in the entire hippocampus of these sections were counted manually and at least four sections were analyzed per animal (from both hemispheres; n ¼ 4 animals per genotype). Limiting analysis to identical regions of hippocampus (and identically sized hippocampi) reduced potential errors from counting different sized regions of hippocampi. Likewise, all images and analysis of PNN distribution in cerebral cortex were obtained from region-and size-matched sagittal sections of wild-type (WT) and littermate mutant mice. PNNs is all layers of motor cortex (mCtx), prefrontal cortex (pfCtx), and primary visual cortex (vCTX) were counted in sagittal sections from mutants and controls (n ¼ 4 animals per genotype). Cortical regions in mutant and control sections were identified blind to the genotype using the Allen Brain Atlas references atlas (http://mouse.brain-map.org/ static/atlas). Quantified data were analyzed with a Student's t-test to determine statistical significance.

In Situ Hybridization
In situ hybridization (ISH) was performed on 16 mm coronal cryosectioned tissues as previously described. The antisense riboprobe generation of full length of adamts4 and syt1 were described previously (Su et al., 2010;Levy et al., 2015). Briefly, riboprobes were synthesized using digoxigenin (DIG)-labeled UTP (Roche, Mannheim, Germany) and the MAXIscript In Vitro Transcription Kit (Ambion, Austin, TX, USA). Probes were hydrolyzed to 500 nt. Sagittal brain sections were prepared and hybridized at 65 C as previously described (Su et al., 2010), and bound riboprobes were detected by horseradish peroxidase (POD)-conjugated anti-DIG antibodies and fluorescent staining with Tyramide Signal Amplification (TSA) systems (PerkinElmer, Shelton, CT, USA). Images were obtained on a Zeiss Axio Imager A2 fluorescent microscope or a Zeiss Examiner Z1 LSM 700 confocal microscope. A minimum of three animals per genotype and age was compared in ISH experiments.

Loss of WFA-Labeled PNNs in the Cerebral Cortex and Hippocampus of Collagen XIX-Deficient Mice
We previously reported that collagen XIX is generated by subsets of interneurons in the mammalian cerebral cortex (Su et al., 2016). To test whether collagen XIX is necessary for the formation or maintenance of PNNs, we assessed the presence of PNNs in several regions of the cerebral cortex of col19a1 À/À mutant brains. PNNs were labeled with biotin-conjugated Wisteria floribunda agglutinin (WFA), a lectin that specifically binds N-acetylgalactosamines (Hartig et al., 1992; Figure 1). In adult wild-type mice, WFA labeled a large set of PNNs that surround the somas and proximal neurites of subsets of interneurons in all regions of cortex (Figure 2(a) to (c)). Similarly, in littermate col19a1 À/À mutants WFA-labeling revealed PNNs in all regions of cortex and these specialized ECM assemblies appeared similar in morphology and size to those in controls (Figure 2(d) to (f) and data not shown), although the intensity of WFA-labeling appeared reduced compared to controls. However, we observed far fewer PNNs in the absence of collagen XIX (Figure 2(a) to (f)). We quantified PNN numbers in three regions of the cerebral cortex (pfCtx, mCtx, and vCtx) and discovered statistically significant reductions in WFAlabeled PNNs in all three regions of col19a1 À/À brains (Figure 2(g) to (i)).
The distribution of WFA-labeled PNNs in mouse brains is far more extensive than the distribution of col19a1-expressing interneurons Hartig et al., 1994;Bruckner et al., 2003;Su et al., 2010;Su et al., 2016). To test whether the loss of PNNs in col19a1 À/À mutants was the result of a local role for collagen XIX or the result of general changes in mice that globally lack collagen XIX, we assessed PNNs in the thalamic reticular nucleus (TRN). TRN is a region of ventral thalamus containing PNN-ensheathed inhibitory projection neurons (Figure 2(j)), but that lacks collagen XIX expression (Su et al., 2016). The number and density of WFA-labeled PNNs in control and col19a1 À/À mutant TRN were indistinguishable (Figure 2(j) to (l)).
Likewise, the number of PNNs in cerebellum (Cblm), another region of mouse brain that lacks col19a1expressing cells (Su et al., 2010), appeared similar control and col19a1 À/À mutant mice (Figure 2(m) to (o)). Taken together, these results suggest a local role for collagen XIX in the assembly or maintenance of WFAlabeled PNNs.
To confirm this hypothesis, we addressed whether the number of WFA-labeled PNNs was altered in the hippocampus of col19a1 À/À mutant mice. The hippocampus, a region within the telencephalon and a component of the limbic system, contains an abundance of PNNs ( Figure  3(a); Bruckner et al., 2003) and interneurons that generate collagen XIX (Su et al., 2010). Indeed, there was a statistically significant reduction in PNN number in col19a1 À/À hippocampi (Figure 3(b) and (c)). The hippocampus can be subdivided into at least five different areas-cornu ammonis (CA) areas 1-3, the dentate gyrus (DG), and the subiculum (Sub). In control mice, we observed prominent WFA-labeled PNNs in all five of these regions, with the highest density in Sub and CA2 (Figure 3(a)). We observed significant reductions in PNN numbers in four of these areas of collagen XIX-deficient hippocampi: CA1, CA3, DG, and Sub (Figure 3(d) to (o)). It is noteworthy that we failed to detect a significant  reduction in WFA-labeling in CA2 in the absence of collagen XIX (Figure 3(a) and (b)). Sparing of PNNs in CA2 of collagen XIX-deficient hippocampi may reflect the distinct molecular architecture and ECM composition of this region (Bruckner et al., 2003;Carstens et al., 2016;Dudek et al., 2016) and the relative lack of collagen XIX expression by cells in this region (Su et al., 2010).

Loss of Acan-Labeled PNNs in col19a1 À/À Cerebral Cortex and Hippocampus
To confirm the loss of PNNs in collagen XIX-deficient cerebral cortex and hippocampus (rather than the loss of WFA-binding sites with these PNNs), we took an alternative approach to label and quantify PNN number; we immunostained sections of control and col19a1 À/À mutant brains with antibodies directed against Acan (Cat315; Matthews et al., 2002;Brooks et al., 2013), a lectican present in most PNNs (Zimmermann and Dours-Zimmermann, 2008). Acan-immunolabeled a similar, but not identical, population of PNNs as WFA in mouse cortex and hippocampus (Figures 4 and 5; Bruckner et al., 2003;Morawski et al., 2012). Similar to the results described above for WFA labeling, we detected a statistically significant reduction in the number of Acan-containing PNNs in pfCtx, mCtx, and vCTX ( Figure 4) and in CA1, CA3, DG, and Sub ( Figure 5) of col19a1 À/À mutants. Based upon the reduced Acan-immunolabeling, we hypothesized that total telencephalic Acan levels were reduced in the absence of collagen XIX. To test this, protein extracts were isolated from the cerebral cortex or hippocampus of postnatal Day 67 (P67) control or col19a1 À/À mutant mice. Western blots revealed a significant reduction in Acan levels in mutant cortex and hippocampus (Figure 6(a) to (d)). Reduced levels of Acan in collagen XIX-deficient mutants were highly specific to regions of mouse brain that normally generate collagen XIX, since levels appeared normal in cerebellum (Figure 6(e) and (f)). Taken together, these results demonstrate that the local loss of collagen XIX leads to significant reductions in both PNNs and PNN constituents in regions of brain that normally generated this unconventional collagen.

Loss of Collagen XIX Leads to Altered Expression of Lectican-Degrading Proteases
The significant reduction in Acan protein led us to explore whether the loss of collagen XIX altered Acan transcription. To test this, RNA was isolated from the cerebral cortex of P56 control and col19a1 À/À mutant mice and levels of acan mRNA were assessed by qPCR. We failed to detect significant alterations in acan mRNA levels in col19a1 À/À mutant cortex (Figure 7(a)) suggesting the reduced Acan levels was not the result of altered transcription.
In fact, the same was true for other components of PNNs. We tested whether the mRNA expression of other integral components of PNNs, including other lecticans (like brevican), tenascin-R, HAPLN family members (HAPLN1-4), or Hyaluronan Synthases (Has1-3) were altered in col19a1 À/À mutant cortex. Previous studies have revealed that partial or complete loss of any of these PNN constituents leads to reduced number or altered morphology of cortical PNNs (Weber et al., 1999;Bruckner et al., 2000;Carulli et al., 2010;Kwok et al., 2010). We found that the loss of Collagen XIX failed to reduce the transcription of mRNAs of any of these PNN constituents (Figure 7), and in several cases increased expression levels were detected (e.g., tenascin-R, HAPLN1, HAPLN3, HAPLN4, and Has2). This increased expression of PNN-associated genes may represent a homeostatic mechanism activated in collagen XIX-deficient mutants to counteract the loss of this collagen and PNNs.
The increase in expression of tenascin-R, HAPLN1, HAPLN3, HAPLN4, and Has2 in col19a1 À/À mutant cortex failed to provide insight into the reduced levels of PNNs and Acan protein in these mutants. However, the increase production of these PNN constituents led us to speculate that collagen loss may also lead to an upregulation of extracellular proteases capable of degrading PNN components. PNN components are degraded by several families of metalloproteinases, including ADAMTSs and MMPs. Many ADAMTSs (including ADAMTS1,4,5,8,9,11,and 15) are classified as lecticanases and not only have a high affinity and activity for cleaving Acan (Abbaszade et al., 1999;Tortorella et al., 1999;Tang, 2001;Porter et al., 2005), but are expressed in mouse cortex (Gottschall and Howell, 2015;Levy et al., 2015). Likewise, many MMPs are expressed in mouse brain and have been shown to cleave PNN constituents (Nagel et al., 2005;Rankin-Gee et al., 2015). We explored whether expression of 12 members of the ADAMTS and MMP families were altered in col19a1 À/À mutant cortex. Eight of these proteases were significantly increased in collagen XIX-deficient cortex (Figure 8), including ADAMTS1, ADAMTS4, ADAMTS5, ADAMTS15, MMP8, MMP9, MMP12, and MMP13.
One of the mRNAs increased in collagen XIXdeficient cortex encodes the most well-studied endogenous Acan-cleaving enzyme in the brain-ADAMTS4 (also called aggrecanase 1; Tang, 2001;Porter et al., 2005). We previously showed that adamts4 mRNA expression was restricted to oligodendrocytes in the mouse cortex and hippocampus (Levy et al., 2015). Here, we discovered that adamts4 mRNA becomes expressed in a subset of neurons in col19a1 À/À mutant cortex and hippocampus (Figure 9(a) to (d)), suggesting that the significant increase in adamts4 mRNA was the result of an upregulation of this gene in a cell-type that normally does not express it. Moreover, we assessed whether protein levels of ADAMTS4 were also altered in col19a1 À/À mutant cortex. ADAMTS4 is produced as a pro-form ($98 kDa) that undergoes cleavage to become a mature, active peptide ($64 kDa) (Tortorella et al., 1999;Vistnes et al., 2014). In mutant and control cortex, we detected similar levels of the pro-form of ADAMTS4; however, increased levels of active ADAMTS4 were present in col19a1 À/À mutant cortex (Figure 9(e)). Densitometry revealed that this increase in active ADAMTS4 was statistically significant (Figure 9(f)). Thus, these results reveal ectopic expression of adamts4 mRNA in mutant brains and an increase in active ADAMTS4 peptides and therefore suggest that increased production and activity of Acan-degrading Figure 7. Alteration of PNN gene transcription in the absence of collagen XIX. Quantitative PCR (qPCR) analysis was performed on RNA extracted from P56 control (Ctl) or col19a1 À/À mutant (KO) cortex. Levels of target gene expression were normalized to levels of gapdh. Data represent mean fold change (AESD) in mRNA expression in KO compared to levels in control samples. * indicates differs from littermate WT by p < 0.05 by Student's t-test. Note that despite reduced levels of Acan protein levels in KO, qPCR revealed Acan transcription was unaltered in KO cortex. PNN genes analyzed: acan ¼ aggrecan; bcan ¼ brevican; tnr ¼ tenascin R; hapln1 ¼ hyaluranonin and proteoglycan binding link protein 1; hapln2 ¼ hyaluranonin and proteoglycan binding link protein 2; hapln3 ¼ hyaluranonin and proteoglycan binding link protein 3; hapln4 ¼ hyaluranonin and proteoglycan binding link protein 4; has1 ¼ hyaluronan synthase 1; has2 ¼ hyaluronan synthase 2; has3 ¼ hyaluronan synthase 3. Figure 6. Reduced levels of aggrecan (Acan) in collagen XIXdeficient cerebral cortex and hippocampus. Western blots demonstrate reduced levels of Acan in protein extracts derived from P67 control (Ctl) or col19a1 À/À mutant (KO) cortex (a), hippocampus (HP; (c)), and cerebellum (Cblm; (e)). Levels of Actin protein were used as loading controls. Quantification of levels of Acan detected by Western blot and normalized to Actin levels. Data shown are mean relative intensities normalized to Actin AE SD (n ¼ 3). * indicates differs from littermate WT by p < 0.05 by Student's t-test. ** indicates differs from littermate WT by p < 0.0001 by Student's t-test.
proteases contribute to the loss of PNNs in collagen XIXdeficient brains.

Discussion
A recent study identified a novel micro-deletion at chromosome 6q12-13 in a cohort of patients with familial schizophrenia, a complex brain disorder characterized by alterations in cognitive function, acquisition of behaviors not observed in healthy individuals (such as hallucinations), and neglect and loss of apathy (Liao et al., 2012). This genomic region contains the coding sequence for collagen XIX. In support of the association of collagen XIX mutation or deletion with schizophrenia, we recently demonstrated that loss of collagen XIX leads to the acquisition of schizophrenia-related behaviors and seizures in mice (Su et al., 2016). These collagen-deficient mutant mice display a striking reduction in the formation of inhibitory axo-somatic synapses in the cerebral cortex and hippocampus (Su et al., 2010;Su et al., 2016), a class of synapse implicated in the mechanisms underlying schizophrenia, seizures, and other neurodevelopmental disorders Lewis et al., 2011;Nakazawa et al., 2012;Wohr et al., 2015). Taken together, the loss of axo-somatic inhibitory synapses in col19a1 À/À mutant mice and schizophrenic patients suggests a possible mechanism of how mutation, reduction or complete loss of collagen XIX may contribute to schizophrenia. However, there are other cellular and molecular alterations in the brains of patients with schizophrenia in addition to the loss of axo-somatic inhibitory synapses. One such alteration is the loss of interneurons (Zhang and Reynolds, 2002;Penschuck et al., 2006;Enwright et al., 2016). Previously we assessed interneuron number and distribution in cerebral cortex and hippocampus of collagen XIX-deficient brains and found no appreciable affect on interneuron number or distribution in the absence of this collagen in mice (Su et al., 2010;Su et al., 2016). Another alteration in patients with schizophrenia is PNN loss (Berretta et al., 2015;Bitanihirwe et al., 2016). Here, we set out to assess whether the loss of collagen XIX altered the number or distribution of PNNs. Our results demonstrate for the first time that the loss of a neuronally expressed collagen results in a significant reduction in the specialized lattice-like ECM that surround cortical and hippocampal interneurons. Therefore, these results identify a second mechanism by which the loss of collagen XIX may contribute to the development of neuropsychiatric disorders, such as schizophrenia.

Does Collagen XIX Play a Direct role in PNN Assembly or Maintenance?
It is possible that interneuron-derived collagen XIX acts as a nucleating factor for PNNs. Outside of the nervous Figure 8. Alteration of protease gene transcription in the absence of collagen XIX. Quantitative PCR (qPCR) analysis was performed on RNA extracted from P56 control (Ctl) or col19a1 À/À mutant (KO) cortex. Levels of protease gene expression were normalized to levels of gapdh. Data represent mean fold change (AESD) in mRNA expression in KO compared to levels in control samples. * indicates differs from littermate WT by p < 0.05 by Student's t-test. Note that eight genes encoding PNN-degrading proteases from the ADAMTS and MMP families of proteases were upregulated in col19a1 À/À mutant (KO) cortex. system, nonfibrillar collagens bind and anchor numerous ECM constituents into supramolecular assemblies (Ricard-Blum, 2011). Collagen XIX is a relatively recently identified, minor collagen (Myers et al., 1997;Sumiyoshi et al., 1997) and therefore the extent to which it binds other ECM components remains to be determined. In addition to its numerous collagenous domains, collagen XIX contains an N-terminal thrombospondin domain which may interact with other components of the brain ECM (Ricard-Blum, 2011). It also contains a C-terminal noncollagenous domain that interacts with a variety of RGD-dependent integrins (Oudart et al., 2015(Oudart et al., , 2016Su et al., 2016) which are known to be expressed by interneurons in the cerebral cortex and hippocampus (Bi et al., 2001;Su et al., 2016). At present it remains unclear whether collagen XIX itself is present in PNNs. While collagen XIX is generated by subsets of cortical and hippocampal interneurons and PNNs preferentially ensheath interneurons in these brain regions, it is important to note that as many as 20 different classes of interneurons exist in these brain regions (Klausberger and Somogyi, 2008). Collagen XIX is generated by classes of interneurons that are defined neurochemically by their expression of calbindin, somatostatin, or neuropeptide Y and only a small fraction of these cells are ensheathed by PNNs (Su et al., 2010;Su et al., 2016). In contrast, PNNs preferentially ensheath classes of telencephalic interneurons that generate parvalbumin  but only a small fraction of these cells appear to generate Collagen XIX (<10% of col19a1-expressing cells coexpress Parv in cerebral cortex; Su et al., 2016). Therefore, while results presented here indicate a local role for collagen XIX in the assembly of PNNs, it is likely that many interneurons ensheathed by PNNs do not generate this collagen. An alternative possibility is that collagen XIX is secreted by one class of interneurons and aggregates around different classes of neighboring cells. This is the case for PNN constituents that are produced by glial cells (John et al., 2006;Giamanco and Figure 9. Ectopic neuronal expression of adamts4 mRNA and increased active ADMATS4 peptides in the absence of collagen XIX. ((a) and (b)) In situ hybridization for adamts4 and syt1 mRNA in adult control or col19a1 À/À mutant cortex (Ctx) and hippocampus (HP). Note the co-expression of adamts4 and syt1 mRNA in neurons in mutant Ctx and HP. ((c) and (d)) In situ hybridization for adamts4 and IHC for NeuN in adult control or col19a1 À/À mutant cortex (Ctx) and hippocampus (HP). Note the ectopic co-expression of adamts4 and NeuN in neurons in mutant Ctx and HP. (e) Western blots of ADAMTS4 in protein extracts from control or col19a1 À/À mutant Ctx. The pro-and active forms of ADAMTS4 are indicated with arrows (pro-form -98 kDa; active form -64 kDa). (f) Quantification of amounts of pro-and active forms of ADAMTS4 in control and col19a1 À/À mutant Ctx. Data shown are mean relative intensities (normalized to actin) AE SD (n ¼ 3). * indicates differ from controls by p < 0.05 by Student's t-test. Scale bar ¼ 20 mm. . And this type of paracrine mechanism appears to be how collagen XIX influences the development of inhibitory axo-somatic synapses in the cerebral cortex and hippocampus (Su et al., 2016). Of course, to test this hypothesis it will be essential to localize collagen XIX protein in PNNs in vivo, something that has not been possible with the current set of molecular tools to interrogate this collagen.
Although collagens may act as nucleating factors in basal laminas, it is entirely possible that the same is not true in PNNs and loss of PNNs may not be the direct result of the absence of this unconventional collagen. This possibility is supported by transcriptional changes of extracellular metalloproteinases known to degrade PNN constituents in col19a1 À/À mutants. Many proteases upregulated in the cerebral cortex of col19a1 À/À mutant mice are known to be generated by cortical interneurons that are coated by PNNs (Levy et al., 2015). So how might the loss of this collagen alter MMP or ADAMTS expression? A likely possibility is that this collagen plays a direct role in regulating transcription factors that act upstream of extracellular protease genes. Certainly the influence of ECM in regulating transcription in this fashion has been well documented outside of the brain (Bissell and Barcellos-Hoff, 1987;Spencer et al., 2007). Alternatively, the loss of collagen XIX is known to lead to molecular or cellular changes in the telencephalon (Su et al., 2010;Su et al., 2016) that may lead to altered protease production. While the number and distribution of PNN coated interneurons does not appear to change in col19a1 À/À mutant brains, synapses onto these cells are altered in these mutants (Su et al., 2016). Perturbing connectivity in the brain is known to lead to alterations in protease production (Zhang et al., 1998;Jourquin et al., 2003), just as perturbing protease expression can alter connectivity (Huntley, 2012). In addition to structural changes in circuitry, the acquisition of behavior phenotypes and seizures in col19a1 À/À mutant mice may lead to altered extracellular protease production. Prolonged seizure induction upregulates MMP and ADAMTS expression and the subsequent loss of PNNs in rodent models (Rankin-Gee et al., 2015). Likewise, an upregulation of ADAMTS15, which degrades Acan and other chondroitin sulfate proteoglycans and which we show is upregulated in cortex of col19a1 À/À mutant mice, is upregulated in the hippocampi of patients suffering from medial lobe epilepsy (Venugopal et al., 2012). Thus, it is possible that altered connectivity and activity in col19a1 À/À mutant mice leads to changes in protease production and a loss of PNNs.

A Role for Collagen XIX in Closing the Critical Period of Neural Circuit Development
We now report two morphological phenotypes in the telencephalon of mice lacking collagen XIX-the loss of axo-somatic, inhibitory synapses onto Parvalbuminexpressing interneurons (Su et al., 2016) and the loss of PNNs from this same class of interneurons. Our data indicate that alterations in the formation of inhibitory axo-somatic synapses appear well before the first PNNs begin to assemble in the developing cerebral cortex or hippocampus (Su et al., 2010;Su et al., 2016). However, it is very likely that these phenotypes are interconnected, since they involve the same cell type and both morphological phenotypes are associated with a variety of complex brain disorders. Relatively late in development and after a period of activity-dependent refinement of neural circuits, PNNs emerge and are thought to form a physical barrier to prevent either the loss of established axosomatic connections or the formation of new connections (Wang and Fawcett, 2012). The emergence of PNNs coincides with the closing of the ''critical period'' in circuit development, a period in which neural activity and experience are critical for shaping neural circuit refinement. To test the link between PNNs, axo-somatic synapse and the critical period of circuit development, previous laboratories have enzymatically degraded PNNs in vCTX by delivering chondroitinaseABC, an enzyme that degrades CSPGs and PNNs. Experimental degradation of PNNs in vCTX led to both the re-emergence of ocular dominance plasticity (Pizzorusso et al., 2002) and enhanced motility of dendritic spines and increased structural and functional neuroplasticity in visual cortical circuits (de Vivo et al., 2013). Taken together with results here, and the temporal expression of collagen XIX in the postnatal brain (Sumiyoshi et al., 1997;Su et al., 2010;Su et al., 2016), we posit that loss of collagen XIX may stall the brain in an immature, plastic state and that its expression may contribute to the closing of the critical period in the developing cortex. While total loss of collagen XIX is detrimental due to the emergence of schizophrenia-related behaviors and seizures, our results raise the intriguing possibility that local manipulation of this collagen may be a useful therapeutic approach to modulate critical periods of cortical plasticity.