Matrix metalloproteinase-3 in the central nervous system: a look on the bright side

Abstract

Matrix metalloproteinases (MMPs) are a large family of proteases involved in many cell-matrix and cell-cell signalling processes through activation, inactivation or release of extracellular matrix (ECM) and non-ECM molecules, such as growth factors and receptors. Uncontrolled MMP activities underlie the pathophysiology of many disorders. Also matrix metalloproteinase-3 (MMP-3) or stromelysin-1 contributes to several pathologies, such as cancer, asthma and rheumatoid arthritis, and has also been associated with neurodegenerative diseases like Alzheimer's disease, Parkinson's disease and multiple sclerosis. However, based on defined MMP spatiotemporal expression patterns, the identification of novel candidate molecular targets and in vitro and in vivo studies, a beneficial role for MMPs in CNS physiology and recovery is emerging. The main purpose of this review is to shed light on the recently identified roles of MMP-3 in normal brain development and in plasticity and regeneration after CNS injury and disease. As such, MMP-3 is correlated with neuronal migration and neurite outgrowth and guidance in the developing CNS and contributes to synaptic plasticity and learning in the adult CNS. Moreover, a strict spatiotemporal MMP-3 up-regulation in the injured or diseased CNS might support remyelination and neuroprotection, as well as genesis and migration of stem cells in the damaged brain.

Abbreviations used

AD

Alzheimer's disease

aNPC

adult neural stem/progenitor cell

BDNF

brain-derived neurotrophic factor

CSPG

chondroitin sulphate proteoglycan

ECM

extracellular matrix

FasL

Fas ligand

IGF

insulin-like growth factor

MMP

matrix metalloproteinase

NG2

neuron/glia-type 2

NGF

nerve growth factor

PD

Parkinson's disease

PNN

perineuronal net

TIMP

tissue inhibitor of metalloproteinases

Matrix metalloproteinases: an overview

Matrix metalloproteinases (MMPs) are proteolytic enzymes that remodel the pericellular environment by degrading all protein constituents of the extracellular matrix (ECM). Besides, they also regulate many cell signalling pathways and homeostatic systems by cleavage and release of various guidance and adhesion molecules, receptors, growth factors, cytokines, etc., through either activation or inactivation. The MMP family, a subgroup of the metzincins, constitutes more than 20 mammalian members, which are all Zn²⁺-dependent endopeptidases. Based on their substrate specificity and domain organization, MMPs are classified into collagenases, gelatinases, stromelysins, membrane-type MMPs, matrilysins and ‘other MMPs’ (Nagase et al. 2006).

MMP activities are kept under tight control. First of all, MMP expression can be regulated at the transcriptional level by growth factors, cytokines, chemokines, hormones, epigenetic processes and cell–cell/cell–ECM interactions. Membrane-trafficking and subsequent release of MMPs at the cell surface can be regulated by SNARE proteins (Kean et al. 2009). MMPs are synthesized as proenzymes with a ‘cysteine switch’, the disruption of the interaction between the cysteine residue in the propeptide and the Zn²⁺ ion in the catalytic site, as a pre-requisite for activation. This disruption can be achieved by organomercurial compounds, heavy metals, denaturating agents or oxidants, as well as through removal of the propeptide by proteases (Van Wart and Birkedal-Hansen 1990). Once activated, MMPs are subjected to inhibition by endogenous inhibitors: α₂-macroglobulin, reversion-inducing cysteine-rich protein with kazal motifs (RECK) and tissue inhibitors of metalloproteinases (TIMPs), the latter also being under tight transcriptional control (Oh et al. 2001; Nagase et al. 2006). The four TIMPs identified so far bind MMPs non-covalently, thereby blocking their activities. As such, MMP activity is largely determined by the MMP/TIMP balance. However, next to gene transcription, zymogen activation and the MMP/TIMP ratio, MMP activity can also be regulated at the less characterized levels of mRNA stability, translational control, for example, acetylation, phosphorylation or S-nitrosylation, (auto)proteolysis and receptor-mediated endocytosis (Chakraborti et al. 2003; Nagase et al. 2006; Clark et al. 2008).

In normal resting adult tissues, in which protease activities are well controlled, MMP levels are quite low. However, when tissue remodelling or cell signalling is required in developing and adult organisms, these proteinases are up-regulated. Consequently, MMPs play an important role in many physiological processes, for example, wound healing, ovulation, blastocyst implantation, bone growth and angiogenesis. Inflammatory stimuli increase MMP levels even more and uncontrolled MMP activity is known to underlie a variety of diseases, for example, tumor invasion and metastasis, rheumatoid arthritis, periodontal disease and atherosclerosis. MMP-over-expression also strengthens blood–brain barrier dysfunction, demyelination, neuroinflammation and neurotoxicity and as such underlies a range of neurological diseases, for example, multiple sclerosis, amyotrophic lateral sclerosis, Alzheimer's (AD) and Parkinson's (PD) disease (Yong 2005; Rosenberg 2009; Klein and Bischoff 2011). Therefore, MMP inhibition has been put forward as a possible therapeutic strategy for treatment of multiple CNS disorders.

Besides a proven detrimental role in neurological disorders, evidence concerning a beneficial contribution of MMPs in key physiological and regenerative events in CNS disorders and injuries is emerging. Furthermore, as neuroinflammation does not only have detrimental consequences, MMP up-regulation is believed to underlie reparative functions in the CNS at well-defined places and time points after an insult (Yong 2010). The currently ongoing identification of MMP targets will undoubtedly provide indications for plausible mechanisms via which MMPs exert these beneficial functions. Among the recently discovered MMP substrates, many ECM proteins, adhesion molecules, chemokines, receptors and growth factors are known to contribute to neuronal processes such as migration, survival, synaptogenesis and plasticity, suggesting a critical role for MMPs in CNS development and regeneration. The use of (non-selective) MMP inhibitors as drug therapy might consequently impede neuroprotective or reparative functions of MMPs.

Therefore, an in-depth knowledge of MMP biology in time and space and the identification of their individual functions and working mechanisms in both physiological and pathophysiological conditions is required.

The stromelysin MMP-3

MMP-3 (EC 3.4.24.17) or stromelysin-1, which belongs together with MMP-10 and MMP-11 to the stromelysin-subgroup, has a rather simple MMP structure with a hemopexin domain attached to the catalytic site by a hinge region. Different molecules, like cytokines, reactive oxygen species, growth factors and cell–cell/cell–ECM interactions can trigger MMP-3 gene transcription, whereas MMP-3 mRNA transcripts are stabilized by phorbol esters and epidermal growth factor (Delany and Brinckerhoff 1992; Kim and Hwang 2011). Once released into the ECM, the inactive MMP-3 pro-enzyme can be activated extracellularly by the plasmin cascade signalling pathway (Nagase et al. 1990). Extracellularly, MMP-3 has a rather broad substrate specificity. Besides degrading multiple ECM proteins, it can also activate growth factors, cleave cell adhesion molecules, chemokines, cytokines and various receptors. In addition, MMP-3 is able to activate pro-MMP-1, -3, -7, -8, -9 and -13 and to hydrolyze some of its upstream activators, such as plasminogen and urokinase-type plasminogen activator (Ogata et al. 1992; Arza et al. 2000; McCawley and Matrisian 2001). Besides this extracellular activity, MMP-3 is known to act intracellularly. As such, MMP-3 is present in dopaminergic neurons, where it can be intracellularly activated by the serine protease HtrA2/Omi (Choi et al. 2008; Shin et al. 2012). Although MMP-3 contains a furin recognition site (Cao et al. 2005) like MMP-11 and MT1-MMP (Pei and Weiss 1996), cleavage by this serine protease does not result in the catalytically active MMP-3 form (Choi et al. 2008). Importantly, secreted MMP-3 can be extracellulary activated and subsequently transported back into the cell, probably via clathrin-dependent endocytosis mechanisms (Traub 2009). In addition, MMP-3 was also discovered intranuclearly, more specifically in the nucleus of hepatocytes (Si-Tayeb et al. 2006) and chondrocytes (Eguchi et al. 2008). The association of MMP-3 with a ras-related nuclear protein (RAN)-binding protein, which is involved in nuclear import, together with nuclear translocation signals in the MMP-3 catalytic site, creates possibilities for efficient transport into the nucleus (Eguchi et al. 2008; Cauwe and Opdenakker 2010). These recent findings and the intracellular and nuclear substrates discovered until now, suggest new functions for MMP-3 in apoptotic processes, transcriptional regulation, protein synthesis and cytoskeletal remodelling (Cauwe and Opdenakker 2010). Both intra- and extracellular activities of MMP-3 can be regulated by TIMPs, among which the TIMP-1/MMP-3 ratio is the best characterized (Kim et al. 2010). An altered balance between TIMPs and MMP-3 is known to impair wound healing (Bullard et al. 1999) and to contribute to the development of, for example, arthritis (Green et al. 2003) and cancer metastasis (Li et al. 1994).

A detrimental role for MMP-3 in the CNS

MMP-3 up-regulation in CNS pathologies implicates its contribution to several neurodegenerative disorders. Indeed, uncontrolled MMP-3 activity seems to aggravate many neurodegenerative disorders by disrupting the blood–brain barrier, by promoting demyelination and apoptosis or by evoking additional inflammatory responses. In the pathological brain, MMP-3 is expressed by injured neurons, oligodendrocytes, astrocytes, pericytes and reactive microglia/macrophages. After ischemia, trauma or infections in the CNS, inflammatory mediators, such as tumor necrosis factor-α and interleukin-1β, induce the transcription factors activator protein-1 and nuclear factor kB, which subsequently bind their analogous sites in the MMP-3 promoter region of activated microglia, thereby inducing pro-MMP-3 transcription. During the injury cascade, other proteases, such as plasmin, are produced and can generate active MMP-3 in the matrix (Rosenberg et al. 2001; Conant and Gottschall 2005; Gurney et al. 2006; Rosenberg 2009).

Up-regulated MMP-3 levels and activity support neuroinflammation as well as apoptosis, two processes which both importantly contribute to neurodegeneration. MMP-3 levels are often increased in neuronal cells after various forms of cellular stress, like endoplasmic reticulum stress, and may ultimately cause apoptosis (Kim et al. 2010). Endoplasmic reticulum stress is involved in the pathogenesis of various neurodegenerative diseases such as PD and AD (Lindholm et al. 2006). MMP-3 up-regulation in dying substantia nigra cells in experimental models of PD (Sung et al. 2005; Kim et al. 2007; Choi et al. 2008) and a decreased cell loss of these nigral neurons after MMP-3 inhibition and in MMP-3-deficient mice (Kim et al. 2007; Choi et al. 2008) has been demonstrated. MMP-3 seems to contribute to the pathophysiology of PD by degrading the neuroprotective protein DJ-1, thereby impeding its antioxidant function, and by intracellulary cleaving the protein α-synuclein, colocalized with MMP-3 in Lewy bodies in stressed dopaminergic cells, thereby leading to the formation of toxic α-synuclein aggregates (Choi et al. 2011a,b). Although the functional relevance for MMP-3 in the pathogenesis of AD still needs to be unraveled, MMP-3 expression has been observed in senile plaques, as well as in plasma and CSF of postmortem brains from AD patients. In addition, altered MMP-3 levels are correlated with altered β-amyloid levels (Deb and Gottschall 1996; Yoshiyama et al. 2000).

MMP-3 has also been linked to multiple sclerosis and demyelination. MMP-3 is able to degrade myelin basic protein and is up-regulated in the brain, prior to the onset of disease in a spontaneous demyelinating mouse model (Chandler et al. 1995; D'Souza et al. 2002). Furthermore, MMP-3 expression is observed in and around lesions and in the blood of multiple sclerosis patients (Maeda and Sobel 1996; Kouwenhoven et al. 2001). An elaborate review describing the involvement of MMP-3 in neurodegeneration was recently published by Kim and Hwang (2011).

A beneficial role for MMP-3 in the developing and adult CNS

However, and despite the fact that MMP-3-deficient mice have no apparent developmental defects (Mudgett et al. 1998), more in depth investigations are becoming to reveal crucial physiological functions of MMP-3 within the developing and recovering CNS. Importantly, MMP-3 cleaves a very wide range of CNS substrates, which associate MMP-3 with pathological processes, but also with diverse physiological functions in the developing and normal adult brain, as well as during recovery/regeneration after CNS insults (Table 1). Moreover, we previously reported mild deficits in balance and motor performance in developing and adult MMP-3-deficient mice (Van Hove et al. 2012). Additional behavioral studies in adult MMP-3-deficient animals also revealed gait abnormalities and showed significantly higher variabilities in stride length and paw angle and an increase in hind base distance in comparison to wild-type littermates, amplifying the contribution of MMP-3 in the normal functioning of the CNS (Fig. 1a–c).

Figure 1. Gait dynamics in adult MMP-3^−/− mice and MMP-3 protein expression patterns in the postnatal mouse cerebellar cortex. Using the DigiGait^™ Imaging System (Mouse specifics, Inc.), mice were imaged ventrally with a digital video camera while running on a motorized transparant treadmill (belt speed = 22 cm/s). Digital paw prints were generated and many gait parameters were determined using the DigiGait^™ software as previously described (Hampton et al. 2004; Vincelette et al. 2007). (a) A similar front base width, but a significantly enlarged hind base width is observed in 8 to 10 week old MMP-3 deficient mice (−/−), as compared with wild-type animals (+/+) (n ≥ 7). (b, c) The coefficient of variation (CV) of stride length (b), calculated from the equation: 100 × standard deviation/mean value, and the variability in paw placement angle (c) were found to be persistently higher in MMP-3^−/− animals, indicating abnormalities in gait pattern as compared with wild-type mice (n ≥ 7). (d) A fluorescent immunohistochemical double labeling for GFAP (green) and MMP-3 (red), performed on cerebella of P8 pups using an antibody against MMP-3 from Epitomics (1908-1, 1 : 200), clearly shows MMP-3 colocalization in Bergmann glial fibers (arrowhead) and other astrocytic processes. (e) Localization of MMP-3 (Epitomics antibody) using a diaminobenzidine (DAB) chromogenic staining confirms these findings. (f) A fluorescent immunostaining for GFAP and MMP-3, using a Santa Cruz antibody (sc-6839, 1 : 200) against MMP-3, shows no colocalization with glial cells, but clear MMP-3 expression in the Purkinje cell layer (PCL), in interneurons of the molecular layer (ML) and the internal granular layer (IGL). (g) A DAB staining for MMP-3 (Santa Cruz antibody) confirms its expression in neuronal cells. (h, i) Western blotting for MMP-3 was performed on equal amounts of proteins from P8 cerebellar extracts (C) and glial cell supernatant (GL) using both antibodies and a 55-kDa rhMMP-3 (R) as standard (R&D WBC015). Both antibodies [Epitomics antibody used at 1 : 200 (h); Santa Cruz antibody used at 1 : 250 (i)] revealed the presence of pro- and active MMP-3 in P8 cerebellar samples. Both antibodies also revealed pro-MMP-3 expression in glial cell supernatant. (i) However, the Santa Cruz antibody also labelled a strong band around 70 kDa in P8 cerebellar samples and a similar but weaker band in the glial cell supernatant. Immunohistochemistry and western blotting were performed as previously described (Van Hove et al. 2012). Scale bar: 100 μm. Data are represented as mean ± SEM (*p < 0.05, **p < 0.01, Student's t-test) (CV, coefficient of variation; var, variability; LF, left front; RF, right front; LH, left hind; RH, right hind.

Table 1. Candidate MMP-3 targets in the CNS. This table provides a categorized list of potential MMP-3 substrates identified in the CNS, with corresponding references listed below

	MMP-3 substrate	Reference
(BDNF, brain-derived neurotrophic factor; HB-EGF, heparin-binding epidermal growth factor; IGFBP-3, insulin-like growth factor binding protein 3; IL-1β, interleukin-1β; MCP, monocyte chemoattractant protein; MMP-3, matrix metalloproteinase-3; NG2, neuron/glia-type 2; NGF, nerve growth factor; PAI-1, plasminogen-activator inhibitor 1; SDF-1α, stromal cell-derived factor 1α; SPARC, secreted protein acidic and rich in cysteine; TNF-α, tumor necrosis factor α; uPA, urokinase-type plasminogen activator; VEGF, vascular endothelial growth factor).
Matrix proteins	Agrin	Sole et al. (2004)
	Aggrecan	Durigova et al. (2011)
	Brevican/neurocan/NG-2 phosphacan/versican	Muir et al. (2002)
	Collagens	Okada et al. (1986)
	Decorin	Imai et al. (1997)
	Elastin	Galloway et al. (1983)
	Entactin	Alexander et al. (1996)
	Laminin/fibronectin	Okada et al. (1986)
	Osteopontin	Agnihotri et al. (2001)
	Perlecan	Whitelock et al. (1996)
	SPARC	Sage et al. (2003)
	Tenascin	Imai et al. (1994)
	Vitronectin	Imai et al. (1995)
Growth factors	Pro-BDNF, pro-NGF	Lee et al. (2001)
	Pro-HB-EGF	Suzuki et al. (1997)
	VEGF	Lee et al. (2005)
Proteases	Plasminogen	Lijnen et al. (1998)
	Pro-MMP-1, -3, -7, -8, -9, -13	Reviewed in McCawley and Matrisian (2001)
	uPA	Ugwu et al. (1998)
Adhesion molecules	E-Cadherin	Fuchs et al. (2005)
Receptors	NMDA-receptor	Pauly et al. (2008)
Cytokines/chemokines	MCP-1, -2, -3	McQuibban et al. (2002)
	Pro-IL-1β	Schonbeck et al. (1998)
	Pro-TNF-α	Gearing et al. (1995)
	SDF-1α	McQuibban et al. (2001)
Others	α2-Antiplasmin	Lijnen et al. (2001)
	α1-Antichymotrypsin	Mast et al. (1991)
	Fas ligand	Matsuno et al. (2001)
	Fibrinogen/fibrin	Bini et al. (1996)
	IGFBP-1, -3	Fowlkes et al. (2004)
	myelin basic protein	D'Souza and Moscarello (2006)
	PAI-1	Lijnen et al. (2000)
	Serum amyloid A	Stix et al. (2001)
	Substance P	Stack et al. (1991)
	t-Kininogen	Sakamoto et al. (1996)

With this review, we intend to stress that MMP-3 actions in the brain are extending beyond their intensively studied detrimental involvement in various neuropathologies. Therefore, we summarize the current knowledge of MMP-3 functions in CNS physiology and recovery/regeneration in the following paragraphs.

MMP-3 in the developing and healthy adult CNS

Expression pattern of MMP-3 in the embryonic CNS

A number of in vitro and in vivo studies showed MMP-3 expression in the embryonic rodent brain and spinal cord. More specifically, immunostainings on embryonic day (E) 14 rat spinal cord neuronal cultures revealed MMP-3 expression in neuronal dendrites, in somata and remarkably also in nuclei after 20 days in culture (Pauly et al. 2008). In the rat embryo at E15, the time point at which axonal outgrowth is intensively ongoing in the central and peripheral nervous system, widespread MMP-3 immunoreactivity was detected in most brain areas and spinal cord commissural axons, but also in peripheral cell bodies and axons, for example, in motor neuron axons, in distal parts of peripheral nerves and their surrounding mesenchymal cells and in dorsal root ganglia containing sensory neuron cell bodies (Nordstrom et al. 1995). Also, studies in E15 mouse brain revealed intense MMP-3 protein staining, especially in neurons of the ventricular zone and cortical plate and in growing axons of the intermediate zone. In situ zymography confirmed MMP-3 activity in these neocortical areas (Gonthier et al. 2007). Using western blotting, RT-PCR and immunohistochemistry on E15 rat cortical cultures, MMP-3 expression was localized in primary cortical neurons, but not in glia (Wetzel et al. 2003). However, a detailed study on E18 whole rat brain isolated cell cultures identified MMP-3 protein in neurons and mature oligodendrocytes, but not in oligodendrocyte precursors, astrocytes and microglia (Sole et al. 2004). In the late gestational fetal rabbit brain (E21 and E26), MMP-3 mRNA was detected in forebrain germinal matrix cells, the source of cortical neurons and glial cells, in most neurons of the diencephalon and in large neurons of the brainstem (Del Bigio and Jacque 1995).

Overall, these data indicate that MMP-3 expression is largely confined to neurons in the embryonic CNS and is only associated with mature oligodendrocytes in the late prenatal period.

Expression pattern of MMP-3 in the postnatal CNS

A widespread distribution of MMP-3 mRNA was detected at various postnatal stages in the rabbit brain (Del Bigio and Jacque 1995). Diencephalic neurons expressed MMP-3 until postnatal day (P)2 and most cells of the cerebral cortex, hippocampus and striatum showed MMP-3 mRNA expression until P10. From P10 on, MMP-3 was found in mature olfactory bulb neurons, in neurons of the facial nucleus, dorsal raphe, gigantocellular region and inferior olive. The cerebellar external and internal granular layer expressed MMP-3 until P20 (Del Bigio and Jacque 1995). Using whole mouse brain extracts, only low and rather temporally unchanged MMP-3 mRNA levels were found at postnatal weeks 1, 3, 5 and 7 (Ulrich et al. 2006). However, real-time PCR analysis in postnatal cerebellar extracts revealed a restricted MMP-3 mRNA pattern with elevated expression levels during the second postnatal week (Van Hove et al. 2012). Using a commercially available and widely used polyclonal antibody for MMP-3 (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA), we showed MMP-3 protein expression in the developing mouse cerebellar cortex. Our immunostaining revealed a change in time from a rather diffuse distribution of MMP-3 in the ECM towards a more restricted staining in cell bodies, with most prominent MMP-3 immunoreactivity in Purkinje cell somata and dendrites and to a lesser extent in GABAergic interneurons and in some granule cells of the external and internal granular layer (Van Hove et al. 2012). This pattern resembles the previously observed MMP-3 immunopattern, obtained after using an antibody from Chemicon, in the developing rat cerebellar cortex, except for an extra diffuse MMP-3 immunopositivity around Bergmann glial fibers in this species (Vaillant et al. 1999). Remarkably, when performing MMP-3 immunostainings on postnatal mouse cerebellar sections using a rather novel commercially available antibody (Epitomics, Inc., Burlingame, CA, USA), MMP-3 immunoreactivity could only be detected in glial cells (see Fig. 1d–g). Western blotting on supernatant of isolated primary mouse cerebellar glial cells, harvested from P8 pups, showed with both antibodies (Santa Cruz and Epitomics), the presence of pro-MMP-3 (55 kDa). Both antibodies also revealed pro-MMP-3 (55 kDa) and active MMP-3 (47 kDa) in extracts of postnatal mouse cerebellum. However, the Santa Cruz antibody additionally labelled a profound band around 70 kDa in the cerebellar samples and a similar but weak band in glial cell supernatant (see Fig. 1h–i). These findings might explain the discrepancy in MMP-3 immunopattern obtained when using both antibodies on mouse postnatal cerebella. However, the western blots clearly reveal MMP-3 expression in glial cells within the postnatal CNS. Of note, additional studies reported glial MMP-3 expression in the postnatal brain. MMP-3 mRNA was detected in subpial and perivascular astroglia as well as in white matter glia, mostly oligodendrocytes, prior to and during myelination of the postnatal rabbit brain (Del Bigio and Jacque 1995). Furthermore, western blotting on cultured primary astrocytes and Schwann cells, isolated respectively from neonatal rat neocortex and sciatic nerves, also revealed MMP-3 expression by these cells (Muir et al. 2002). Overall, these findings indicate the presence of MMP-3 in neurons as well as glial cells in the postnatal CNS.

Expression pattern of MMP-3 in the adult CNS

In the adult healthy CNS, very low to undetectable MMP-3 (m)RNA and protein levels are observed (Pagenstecher et al. 1998; Rosenberg et al. 2001; Ulrich et al. 2005; Li et al. 2009). In the adult rabbit brain, MMP-3 mRNA expression is confined to the dorsal raphe (Del Bigio and Jacque 1995). Western blotting revealed pro-MMP-3 levels in adult rat and mouse brain or isolated hippocampus, but active MMP-3 levels were low to undetectable (Meighan et al. 2006; Wright et al. 2006; our findings). Immunohistochemistry in the adult rat cerebellum showed MMP-3 protein expression in interneurons of the molecular layer, Purkinje cell somata and granule cells in the internal granular layer (Vaillant et al. 1999). MMP-3 immunoreactivity was also detected in the adult rat hypothalamo-neurohypophysial system, more specifically in a punctate pattern at the dendrites and terminals of arginine-vasopressin expressing magnocellular neurons (Miyata et al. 2005). In the healthy elderly human brain (age range from 71 to 93 years), diverse studies showed perineuronal, neuronal and neuropil MMP-3 labelling in the frontal cortex, but MMP-3 was not found in the parietal cortex or hippocampus (Yoshiyama et al. 2000; Baig et al. 2008). MMP-3 expression was also detected in microvessels and perivascular microglia in the healthy human brain (Maeda and Sobel 1996). Of note, these MMP-3 immunostainings do not distinguish between pro- and mature MMP-3 forms. However, active MMP-3 could be detected in CSF of healthy subjects (Candelario-Jalil et al. 2011). Thus, the reported findings indicate that MMP-3 expression is largely confined to neurons and perivascular microglia in the healthy CNS.

In general, the variations/discrepancies reported in MMP-3 expression patterns observed in the embryonic, postnatal or adult CNS, might result from the use of various antibodies targeting different epitopes and/or might be caused by distinct MMP-3 post-translational modifications and interactions, together with a time-, place- and species-specific localization. Future localization and activity studies, using diverse antibodies, MMP-3-enhanced green fluorescent protein (eGFP) reporter animals or improved activity assays, should provide detailed MMP-3 spatiotemporal expression and activity patterns, thereby enabling its functional characterization in various regions of the CNS.

MMP-3 in neuronal patterning and axon outgrowth during embryonic and postnatal CNS development

The previously described spatiotemporal expression patterns suggest a role for MMP-3 during CNS development. Furthermore, an involvement of MMP-3 in brain development can also be deduced from the large list of MMP-3 targets present in the developing brain and known to be associated with one or more brain developmental processes, such as neurite outgrowth, neuronal migration, survival, synaptogenesis and/or myelinogenesis (Tables 1, 2) (also see Fig. 2).

Figure 2. Beneficial roles of MMP-3 in the developing and adult CNS. This scheme depicts the various cell types expressing MMP-3 and its contribution to important processes in the developing and adult CNS. Previous expression studies in the developing and healthy adult CNS and during recovery/regeneration in the adult CNS show MMP-3 expression in neurons, oligodendrocytes, astrocytes and perivascular microglia. In the developing CNS, MMP-3 contributes to neurite outgrowth, axon invasiveness, axon guidance and neuronal migration (shown in green). In the healthy adult CNS, MMP-3 is known to support synaptic plasticity, learning and memory (shown in red). Up-regulated MMP-3 levels might underlie recovery/regeneration after adult CNS injury by promoting remyelination, neuroprotection and neural stem cell genesis and migration and might contribute to axonal regeneration and synaptic plasticity in the damaged brain (shown in blue).

Table 2. Cellular and nuclear MMP-3 substrates. This table lists the intracellular and intranuclear MMP-3 substrates known to date, with corresponding references

MMP-3 substrate	Reference
Actin-related protein (Arp) 2/3 complex subunits	Cauwe et al. (2009)
α-Synuclein	Sung et al. (2005)
DJ-1	Choi et al. (2011a)
Gelsolin	Park et al. (2006)
Histidyl-tRNA synthetase	Cauwe et al. (2009)
Nucleolin	Cauwe et al. (2009)
TRENDIC	Eguchi et al. (2008)
Tubulin α/β	Cauwe et al. (2009)

One study demonstrated a correlation between MMP-3 mRNA expression and the time course of neurite extension in PC12 cells treated with nerve growth factor (NGF) (Machida et al. 1989). Additional in vitro experiments revealed MMP-3 immunoreactivity in the growth cones of NGF treated PC12 cells and a reduced ability of neurites from PC12 cells, expressing antisense MMP-3 mRNA, to penetrate a basal lamina-like matrix (Nordstrom et al. 1995). Further in vitro experiments on cultured E15 mouse cortical neurons showed increased extracellular MMP-3 protein and activity levels after Sema3C treatment and a deviating axonal orientation towards the chemoattractant Sema3C in a growth cone turning assay in the presence of a specific MMP-3 inhibitor (Gonthier et al. 2007). Together, these data clearly suggest an involvement of MMP-3 in axonal extension, growth cone invasiveness and axon guidance.

Recently, MMP-3 was shown to be functionally involved in the postnatal development of the mouse cerebellar cortex. Detailed phenotyping of cerebellar development in MMP-3 deficient mice revealed a protracted tangential granule cell migration and a delayed granule cell radial migration, together with a disturbed Purkinje cell dendritogenesis. These deficits were also associated with a subsequent retarded GABA-ergic interneuron migration. Importantly, a reduced size of the Purkinje cell dendritic trees was shown in cerebella of adult MMP-3^−/− mice and might, in part, underlie the aberrant motor performance observed in these animals (Van Hove et al. 2012). MMP-3 might possibly exert its effects on postnatal cerebellar development via activation of MMP-9 (Ogata et al. 1992), known or suggested to be important for granule cell migration, apoptosis and axonal outgrowth in the cerebellum (Vaillant et al. 2003). However, although the mRNA and protein expression profiles of MMP-3 and MMP-9 in the developing cerebellum are quite comparable, MMP-3 deficiency does not phenocopy the findings observed in MMP-9 deficient cerebellar cortices (Vaillant et al. 1999, 2003), supporting the idea that additional molecules and pathways contribute.

Besides its effect on neurite outgrowth and neuronal migration, MMP-3 deficiency also resulted in a delayed or abrogated synapse formation between granule cells and Purkinje cells in the mouse postnatal cerebellar cortex (Van Hove et al. 2012). Remarkably, during cerebellar development, MMP-3 mRNA and protein expression levels are high when synaptogenesis between granule cells and Purkinje cell dendrites peaks (Vaillant et al. 1999; Van Hove et al. 2012). MMP-3 might influence the formation and function of synapses in the CNS by cleaving molecules which are directly or indirectly involved in synaptic functioning (Ethell and Ethell 2007), for example, laminin (Okada et al. 1986), tenascin (Imai et al. 1994) and pro-brain-derived neurotrophic factor (pro-BDNF) (Lee et al. 2001). Although a functional implication of MMP-3 in developmental synaptogenesis remains largely elusive, this hypothesis can be strengthened by several studies demonstrating a supporting role of MMP-3 in synapse remodelling after CNS injuries (see below).

MMP-3 has also been suggested to contribute to postnatal myelinogenesis (Del Bigio and Jacque 1995; Muir et al. 2002). This process requires matrix remodelling and ECM modifications for proper adhesion as well as an increased availability of growth factors for oligodendrocyte maturation and survival, processes which can be achieved by MMPs. It has been shown that MMP-9 is expressed on growing tips of oligodendrocytes in vitro and during initiation of myelin formation in the mouse optic nerve, thereby, MMP-9 is known to facilitate process outgrowth and oligodendrocyte maturation (Oh et al. 1999; Larsen et al. 2006). Moreover, MMP-9, which can be activated by MMP-3, might regulate developmental myelination by affecting insulin-like growth factor (IGF) bioavailability (Larsen et al. 2006). MMP-3 expression has also been observed in oligodendrocytes and Schwann cells (Del Bigio and Jacque 1995; Muir et al. 2002) and is known to promote the release of IGF by degrading IGF/IGF-binding protein complexes (Fowlkes et al. 1994, 2004). As a role for MMP-3 in developmental myelination was only suggested based on expression patterns and myelination-associated MMP-3 targets, there is an obvious need for more in-depth research on the involvement of MMP-3 in myelin formation. Although all the above findings already indicate the importance of MMP-3 in CNS development, further functional studies are required, complemented with the search for MMP-3 target molecules and mechanisms underlying the observations during CNS development.

MMP-3 in plasticity, learning and memory in the adult healthy brain

A role for MMPs in synapse formation, maturation, plasticity and long-term potentiation has been demonstrated or suggested, in part based on the identification of MMP substrates, known to support these processes (Nagy et al. 2006; Ethell and Ethell 2007; Wang et al. 2008; Bajor et al. 2012) (Table 1).

In the adult mammalian CNS, most neuronal somata and their dendrites are enwrapped by perineuronal nets (PNNs). These PNNs consist of several ECM molecules, including the chondroitin sulphate proteoglycans (CSPGs) aggrecan, neurocan, brevican, versican, phosphacan, and their binding partners, tenascin-R, hyaluronan and link proteins. Cleavage of these PNNs is necessary to allow synaptic plasticity in the adult CNS (Carulli et al. 2006) and an involvement of MMPs in the turnover of these PNN molecules, thereby stimulating plasticity, is not unlikely (Howell and Gottschall 2012). It was already shown that MMP-3 can cleave all of these CSPGs (Muir et al. 2002; Durigova et al. 2011) and the observed perineuronal MMP-3 protein localization indicates an association of MMP-3 with PNN remodelling (Baig et al. 2008). In addition, murine slices of ventral hippocampus showed an increase in amplitude of the field excitatory post-synaptic potential after exposure to MMP-3. Together with a prominent disruption of PNNs in slices exposed to a specific MMP-3 inhibitor, these data suggest a possible interaction between MMP-3 and PNN integrity in the hippocampus (Gordon D., Greene J.R.T. and Betmouni S., unpublished observations).

Studies performed in the PNS showed active MMP-3 at the neuromuscular junction and reduced MMP-3 activity following denervation and after blocking nerve activity at the neuromuscular junction. Moreover, MMP-3 has been associated with synaptic sprouting in the PNS and is able to cleave synaptic basal lamina molecules, including agrin, a transmembrane protein involved in synapse differentiation and synaptic transmission (VanSaun et al. 2003, 2007; Werle and VanSaun 2003; Sole et al. 2004). Although MMP-3 is also able to cleave the brain agrin isoform (Sole et al. 2004), a functional correlation between MMP-3 and agrin in brain plasticity has not been determined yet.

Most studies investigating a possible involvement of MMP-3 in CNS synaptic plasticity are performed in the hippocampus. A transient increase in hippocampal MMP-3 mRNA and protein levels has been observed during the active phase of spatial learning in a Morris water maze test (Meighan et al. 2006). These changes in MMP-3 levels, possibly as a consequence of NMDA receptor activation, might consequently contribute to activity-dependent changes in synapse morphology and physiology. Importantly, NMDA-induced shedding of the synaptic intercellular adhesion molecule-5 (ICAM-5) is associated with dendritic spine maturation and might be regulated, at least in part, by MMP-3 (Tian et al. 2007; Conant et al. 2010). Next to disrupting ECM molecules, plausible actions of MMP-3 in changing synaptic morphology and efficacy during learning, include (in)activation and/or release of non-ECM molecules, interruption of ECM-cell adhesion interactions and modification of the cytoskeleton (Meighan et al. 2006). As shown in Table 2, MMP-3 has indeed intracellular substrates known to be involved in cytoskeletal organization. An influence of MMP-3 on synaptic plasticity and memory consolidation was further elucidated during passive avoidance conditioning, a hippocampus-dependent associative memory task, where MMP-3 levels transiently augmented (Olson et al. 2008). In addition, learning deficits were observed after administration of a MMP-3 inhibitor prior to passive avoidance training, suggesting a causal relationship between learning-induced elevated MMP-3 expression in the hippocampus and associative memory formation. MMP-3 levels were also elevated in the hippocampus, prefrontal and piriform cortex after habituation of the head-shake response. This increase in MMP-3 expression, accompanied by elevated MMP-9 activity, indicates that both MMPs could mediate changes in neural plasticity following habituation (Wright et al. 2006). Also, the observed punctate MMP-3 immunopattern in magnocellular neuronal dendrites and terminals of the hypothalamo-neurohypophysial system, which is probably corresponding to neurosecretory granules, suggests an involvement of MMP-3 in inducing structural plasticity (Miyata et al. 2005). Importantly, synaptic plasticity implies structural and functional alterations at the post-synaptic membrane (Wheal et al. 1998) and MMP-3 can contribute to the necessary structural alterations of clustered post-synaptic NMDA receptors through activity-dependent shedding of the NMDA receptor NR1 subunit, as shown in cultured spinal cord neurons (Pauly et al. 2008). Clearly, multiple reports have indicated MMP-3 as a regulator of synaptic plasticity and learning in the CNS and future studies will undoubtedly reveal additional involvements in various brain regions in defined conditions.

MMP-3 in the adult pathological CNS

MMP-3 in neuroprotection

Through proteolytic shedding of receptors, ligands and signalling molecules at the cell surface, MMP activity might mediate cell survival/cell death signalling pathways in various cell types. One route initiating the caspase activation cascade starts after binding of the membrane protein Fas ligand (FasL) to the cell surface death receptor Fas, which are both up-regulated following injury or cellular stress in the CNS (Felderhoff-Mueser et al. 2000; Qiu et al. 2002). In vitro experiments with human L5178Y FasL transfected cells showed the ability of MMP-3 to cleave FasL from the cell surface into its soluble form (Matsuno et al. 2001). Furthermore, in cultured cortical neurons, treated with the chemotherapeutic drug Doxorubicin or subjected to oxygen-glucose deprivation, both known to evoke Fas-mediated apoptosis, TIMP-3 was required for apoptotic death while addition of exogenous active MMP-3 promoted neuroprotection by reducing Fas–FasL interactions (Wetzel et al. 2008). As the TIMP-3/MMP-3 balance seems to facilitate neuronal apoptosis through net MMP-3 inhibition, FasL ectodomain shedding by addition of MMP-3 or removal of TIMP-3 might therefore support neuroprotection (Wetzel et al. 2003, 2008).

Alternatively, MMP-3 might also modulate neuroprotection via increasing the bioavailability of IGF-1 through degradation of IGF/IGF-binding protein complexes (Fowlkes et al. 1994, 2004) and by releasing and cleaving pro-NGF and pro-BDNF (Lee et al. 2001; Cunningham et al. 2005). Of note, mature NGF and BDNF promote cell survival by mainly activating Trk receptors while their proform rather mediates cell apoptosis via activation of the p75^NTR -receptor (Lee et al. 2001; Teng et al. 2005). Overall, although there is evidence that MMP-3 up-regulation might support neuroprotection, still a lot remains to be discovered about its involvement in this process.

MMP-3 in post-injury axonal growth

Although a role of MMP-3 in axonal regeneration remains largely elusive, some reports do already provide a reasonable indication for its involvement in such processes, albeit in the PNS. After sciatic nerve crush, MMP-3 is up-regulated by regenerating peripheral nerve fibers (Demestre et al. 2004; Shubayev and Myers 2004). As developmental studies showed MMP-3 expression in growing axons and axonal growth cones as well as an involvement of MMP-3 in neurite outgrowth and guidance (Nordstrom et al. 1995; Gonthier et al. 2007), a possible role of MMP-3 in adult CNS axonal outgrowth can be hypothetized. Furthermore, as MMP-3 is able to digest all known axon-inhibitory CSPGs in the glial scar, e.g. phosphacan, neurocan, versican, neuron/glia-type 2 (NG2) and brevican, up-regulated MMP-3 levels might create a permissive environment promoting neurite outgrowth by removing the growth-inhibitory glial scar (Muir et al. 2002; Pizzi and Crowe 2007). Of note, removal of the inhibitory environment is not sufficient to promote axonal regrowth; also regeneration has to be promoted! Regenerative properties could be enhanced via production or activity of intrinsic growth factors such as NGF (Silver and Miller 2004), a process to which MMP-3 might well contribute. Overall, future research on MMP-3 in axonal growth should be encouraged as this MMP seems to be a potential candidate protease necessary to overcome the inhibitory environment and to support axonal regrowth after CNS injuries.

MMP-3 in post-injury synapse formation

Degradation of CSPGs might not only support axonal growth, but also synaptic reorganization. Two injury models in rats, namely the traumatic brain injury model, characterized by deafferentiation of target neurons which induces reactive synaptogenesis and functional recovery (Steward 1989), and the combined traumatic brain injury-bilateral entorhinal cortical lesion model, generating additional neuroexcitation but less synaptic recovery, were applied to study changes in MMP-3 levels associated with trauma-induced synaptogenesis. MMP-3 mRNA, protein and activity levels were elevated, not only during the degenerative period, but also during the later regenerative phase characterized by rapid synaptic plasticity. This injury-induced up-regulated MMP-3 expression was localized predominantly in reactive astrocytes within the denervated neuropil, likely participating in removing debris and reshaping the extracellular environment for efficient synapse reorganization (Kim et al. 2005; Falo et al. 2006). Importantly, the spatially and temporally coordinated MMP-3 expression and activity levels correlate with increased expression of its target ECM substrates, molecules necessary to prepare the local environment for reactive synaptogenesis (Deller et al. 2001; Dityatev and Schachner 2003).

As mentioned before, MMP-3 supports synaptic plasticity and learning in the normal adult CNS. Consequently, it can be hypothetized that the MMP-3 up-regulation, described above, might underlie post-injury synaptic plasticity. Future investigations addressing these possibilities might create valuable findings to overcome the limited synaptic reorganization after CNS injury.

MMP-3 in remyelination

An involvement of MMP-3 in remyelination during the regenerative period after CNS injury was demonstrated for the first time in the murine cuprizone-induced demyelination model. More specifically, not only during the early demyelination stage, but also during the stage of remyelination, MMP-3 was highly expressed in astrocytes of the corpus callosum. Although the exact mechanisms remain elusive, MMP-3 might be involved in remyelination via its known involvement in releasing IGF, which is essential for proliferation and differentiation of myelin-forming cells (Dubois-Dalcq and Murray 2000; McCawley and Matrisian 2001; Fowlkes et al. 2004). In addition, MMP-3 might support remyelination by removing and cleaving myelin debris, which inhibits oligodendrocyte precursor cell differentiation (D'Souza and Moscarello 2006; Kotter et al. 2006; Skuljec et al. 2011). Also, accumulation of NG2 after axonal injury not only creates an inhibitory environment for axonal growth, but is also unfavorable for proper maturation of oligodendrocytes (Larsen et al. 2003). Interestingly, NG2 is one of the CSPGs known to be cleaved by MMP-3, and up-regulated MMP-9 expression facilitates remyelination by removal of NG2 (Muir et al. 2002; Larsen et al. 2003).

All the above findings already provide a promising contribution of MMP-3 in CNS remyelination and warrant future in depth investigations.

MMP-3 in adult neurogenesis and migration

In discrete regions of the normal adult brain, more exactly in the subventricular and subgranular zone of the hippocampus, neurogenesis remains ongoing in the adult stage. In response to focal ischemic injury in the brain, proliferation of adult neural stem/progenitor cells (aNPCs) is enhanced in the subventricular zone and followed by migration of the neuroblasts towards the ischemic area (Kokaia and Lindvall 2003). A recent in vitro study revealed an increase in MMP-3 levels in differentiated mouse aNPCs when migrating in response to injury-induced chemokines. Importantly, blocking MMP-3 expression with specific siRNAs inhibited both aNPC differentiation and chemokine-induced neuroblast migration (Barkho et al. 2008). These data were confirmed in a rodent focal ischemia model, showing migrating neuroblasts expressing MMP-3 mRNA, thereby probably allowing these cells to mediate a neurogenic response after ischemic insults (Barkho et al. 2008).

Conclusion

Until now, research on MMP-3 in the CNS has mainly focused on the involvement of this protease in pathological conditions. However, in this review, we consider the increasing number of reports that also indicate a beneficial contribution of MMP-3 in developmental events as well as in recovery processes in the CNS. The complex role of MMP-3 in amplifying or attenuating CNS disorders/injuries depends on when and where MMP-3 is exactly up-regulated, combined with the availability of its potential substrates.

Although a lot remains to be discovered, the existing literature suggests that MMP-3 expression is strictly regulated and involved in a wide variety of neuronal developmental processes like neurite outgrowth, migration and myelination. In the normal adult brain, elevated levels of MMP-3 are shown to contribute to activity-related plasticity and learning. Finally, in the injured or diseased CNS, a time-, space- and cell type-specific coordinated MMP-3 expression and activity seems to create dual functions for this enzyme: in the acute phase after injury, MMP-3 might be associated with CNS degeneration, while in the later regenerative stages, up-regulated MMP-3 levels seem to contribute to axonal sprouting, remyelination and reactive synaptogenesis (as summarized in Fig. 2).

In conclusion, the current data available today already indicate advantageous and critical roles for this protease in the brain, supporting the need to reappraise the use of (non) selective MMP inhibitors as drug therapy in neurological as well as non-neurological disorders. Combined use of proteomic and genetic approaches, using conditional and cell-specific transgenic animals, along with the generation of specific synthetic and/or protein-based inhibitors for MMP-3, will be important to gain more insight in the exact MMP-3 working mechanisms and the substrates involved, and should enable the use of more specifically targeted therapeutic strategies.

Acknowledgements

The authors have no conflict of interest. This work was supported by grants from the Flemish Institute for the promotion of Scientific Research (IWT), the Research Council of KU Leuven and the Research Foundation Flanders (FWO).