Matrix Metalloproteinase Dependent Cleavage of Cell Adhesion Molecules in the Pathogenesis of CNS Dysfunction with HIV and Methamphetamine

Abstract

Physiologically appropriate levels of matrix metalloproteinases (MMPs) are likely important to varied aspects of CNS function. In particular, these enzymes may contribute to neuronal activity dependent synaptic plasticity and to cell mobility in processes including stem cell migration and immune surveillance. Levels of MMPs may, however, be substantially increased in the setting of HIV infection with methamphetamine abuse. Elevated MMP levels might in turn influence integrity of the blood brain barrier, as has been demonstrated in published work. Herein we suggest that elevated levels of MMPs can also contribute to microglial activation as well as neuronal and synaptic injury through a mechanism that involves cleavage of specific cell and synaptic adhesion molecules.

I. INTRODUCTION

Matrix metalloproteinases (MMPs) are a family of zinc dependent endoproteases with 24 expressed in humans [1, 2]. The majority are released from cells in a regulated manner. These soluble MMPs typically contain a pro- domain, a catalytic domain and a hemopexin-like domain which can function in binding interactions. Chromosomal loci vary though a cluster is localized to chromosome 11 [3, 4]. Molecular weights also vary and range from approximately 20 kDa for active MMP-7, which lacks a hemopexin-like domain, to 92 kDa for the gelatinase MMP-9. Following their release, MMPs are activated by removal of their pro domain or by factors that influence tertiary structure including oxidation and nitrosylation [5]. Reductions in soluble MMP activity can also be affected by interactions with endogenously expressed inhibitors known as tissue inhibitors of metalloproteinases (TIMPs)[6]. The MMP family also includes membrane-type MMPs (MT-MMPs), which possess a transmembrane domain as do the related “a disintegrin and metalloproteinase[s]” or ADAMs. The latter may be activated by intracellular signaling molecules including protein kinase C [7].

MMPs were named for their ability to process proteins of the extracellular matrix but are now appreciated to act on a variety of soluble molecules and cell surface receptors as well [8, 9]. These enzymes are expressed in the brain by varied cell types including neurons and glial cells [10, 11], and their expression, release and/or activity may be greatly enhanced in the setting of CNS infection or injury. MMPs typically have AP-1 and NF-κB transcription factor binding sites within their promoters, and their expression may also be regulated by hypoxia inducible transcription factor signaling [10]. MMPs have also been localized to preformed vesicular stores [12] from which their release may be facilitated by stimuli that increase intracellular calcium [13]. In addition, activation of released MMPs may be facilitated by injury-associated stimuli including serine proteases [14] and nitric oxide [5].

MMPs can overlap in terms of substrate specificity [15], though specificity for select substrates has been observed and the efficacy with which individual MMPs can act on specific substrates can vary. For example, one study noted relative selectivity of MMP-1 in terms of MMP dependent activation of the thrombin receptor protease activated receptor-1[16].

Studies of MMPs in the CNS have generally focused on their ability to target extracellular matrix proteins of the blood brain barrier [BBB]. Inflammatory cells may release MMPs to facilitate their transmigration through the blood brain barrier, and high levels of CNS derived MMPs likely diminish BBB integrity as well. Consistent with this, tight junction proteins are targeted by MMPs in the background of stroke [17], and CNS infiltration of inflammatory cells is reduced in MMP knockout animals [18]. MMPs have also been shown to target myelin proteins, and MMP inhibitors have thus been considered for the treatment of specific demyelinating disorders [19]. This review will, however, focus on another important target of MMP activity- cell adhesion molecules [CAMs].

CAMs are ideally positioned to be cleaved before MMP activity is squelched by endogenous soluble inhibitors. In addition, while basal shedding of CAMs occurs, CAM cleavage may be substantially increased in the setting of HIV infection and psychostimulant abuse. Moreover, as will be discussed, shed CAMs may interact with integrins to profoundly influence neuronal and glial function.

II. MMP LEVELS MAY BE SUBSTANTIALLY ELEVATED IN THE SETTING OF HIV INFECTION AND PSYCHOSTIMULANT USE

MMPs are elevated in association with HIV associated neurological disorders (HAND) [20–22]. CSF levels of MMP-2 and -9 in particular are increased. Moreover, HAND relevant stimuli including HIV proteins and TNF-α, can increase MMP release from CNS derived cells [20–23]. For example, injection of gp120 into rat caudate-putamen leads to increases in MMP-2 and -9 [24]. Stimulation of cultured cells with viral proteins is also linked to increases in MMP-2 and -9 [23, 25]. Moreover, HIV-1 Tat stimulates MMP-1 and -2 release from cultured neural cells [23] as well as MMP-9 from monocytes [26].

Similarly, methamphetamine (MA) has been associated with increase MMP levels in the CNS [27]. For example, chronic MA (2 mg/kg/day for five days) is associated with increased MMP-2 and -9 protein in the frontal cortex and nucleus accumbens of rats, and an acute high dose of MA (40 mg/kg) is followed by increased mRNA expression of MMP-9 in the CNS of mice [28]. We have also observed that acute MA increases MMP-9 levels in murine striatum and hippocampus [29].

MA is thought to increase extracellular monoamines [dopamine, serotonin and norepinephrine] through mechanisms that include a redistribution from synaptic vesicles to the cytoplasm and reverse flux through cell surface transporters [30, 31]. MA can also lead to an increase in extracellular glutamate levels [32]. MA associated increases in neurotransmitters may in turn lead to transmitter receptor dependent increased gene transcription. This mechanism is supported by published studies in which catecholamines have been shown to increase MMP expression [33–36].

Since recent studies also suggest that MMPs exist in perisynaptic vesicular stores [12], and that release of vesicular MMPs may be soluble NSF attachment protein receptor (SNARE) dependent [13], MA dependent changes in neurotransmitter levels might also stimulate SNARE dependent release of MMPs from vesicular stores. Consistent with this, a glutamate receptor agonist is associated with rapid MMP dependent substrate cleavage [37].

III. SPECIFIC CAMS, WITH KNOWN INTEGRIN BINDING MOTIFS, REPRESENT AN IMPORTANT CLASS OF MMP SUBSTRATES

CAMs represent an important class of MMP substrates, and CAM cleavage has been extensively investigated in the setting of cell migration and cancer biology. Though CAM cleavage has been less well studied in the CNS, the brain is rich in these molecules. Many are highly expressed in areas critical to brain function, including the neuronal synapse. Neuronal and glial CAMs include specific cadherin [Cdh] family members as well as CAMs with Ig-like domains. Among the latter are intercellular adhesion molecule- and -5 (ICAM-1 and ICAM-5), neural cell adhesion molecule (NCAM), and vascular cell adhesion molecule (VCAM). Shedding of these substrates can occur to a substantial extent, in that soluble shed forms of many CAMs can be detected in blood, brain, and CSF [38–41]. Shed forms of select CAMs including ICAM-5 have been described in control specimens, consistent with baseline shedding [29, 40, 41], and levels in blood or CSF are increased with CNS inflammation or other forms of neurological disease [38–41]. For a list of CAMs that are expressed in the CNS and targeted by MMPs, please see Table 1.

Table 1.

Partial List of CNS CAMs Processed by Metalloproteinases

Cell Adhesion Molecule	Cell Types for which Expression has been Reported	Metalloproteinase(s) that Mediate Shedding	Reference(s)
ICAM-1	Astrocytes, oligodendrocytes, leukocytes, endothelial cells	ADAM-17	[86–90]
ICAM-5	Glutamatergic neurons of the telencephalon	MMP-2,-3,-7,-9	[37, 75]
NCAM	Neurons, astrocytes, oligodendrocytes, microglia, endothelium	ADAM protease	[91–95]
L1-CAM	Neurons, oligodendrocytes, endothelial cells	ADAM-10, ADAM-17	[96–99]
VCAM	Neurons, astrocytes, microglia, endothelial cells	ADAM-9	[100–104]
N-Cadherin	Astrocytes, oligodendrocytes	ADAM-10, MMP-7	[7, 105–108]
E-Cadherin	Neurons and endothelial cells	MMP-7, ADAM-10, ADAM-15	[109–114]
VE-Cadherin	Neuronal stem cells, astrocytes and endothelial cells	MMP-7	[115]
Nectin-1	Neurons	ADAM-10	[116–118]
Syndecan-1	Neurons and astrocytes [expression upreulated with injury], endothelial cells	MT1-MMP, MMP-7, MMP-9	[119–125]
SIRP-1α	Neurons, astrocytes, microglia, endothelial cells	Unknown metalloproteinase[s]	[126–130]

CAM cleavage may have important consequences with respect to CNS function. Loss of adhesive contacts may impair cell or synaptic signaling or stability, with effects on cell survival. In addition, MMP dependent cleavage is often followed by intramembranous proteolysis mediated by gamma secretase, leading to the generation of specific CAM intracellular domains (ICDs). While these may be degraded by the proteosome system, some have the potential to alter cellular transcription. An example is provided by ICDs generated following cleavage of N-cadherin. Ectodomain shedding of N-cadherin is followed by intramembranous proteolysis which generates a C-terminal fragment known as CTF-2. This fragment binds the transcription factor CREB-binding protein (CBP) and promotes its proteosomal degradation so that CREB-mediated transcription is repressed [42].

A third consequence of CAM shedding relates to the generation of shed N-terminal domains that are bioactive. As mentioned, these shed domains are stable enough to be detectable in CSF. As we will discuss, many can also interact with integrins that play an important role in microglial activation and synaptic function.

Integrins are heterodimeric cell surface proteins containing two distinct subunits. These dimers represent an especially important receptor class for shed CAMs. For example, nectin [43], E-cadherin, VCAM (VLA4), ICAM-1 (LFA-1 and Mac-1), and ICAM-5 (LFA-1 and β₁) [29, 44, 45] can all interact with integrins in cis or trans. Members of the immunoglobulin superfamily such as ICAMs and VCAMs in particular have been referred to as a main class of integrin binding ligands [45]. The binding of full length CAMs to integrins might promote stable cell-cell interactions, while the binding of shed CAM ectodomains to unengaged integrins might allow for rapid changes in integrin signaling. In studies focused on ICAM-5, both full length and shed forms were found to co-immunoprecipitate with β₁ integrins. In addition, we have observed that soluble ICAM-5 stimulates integrin dependent phosphorylation of cofilin in neuronal cells, an event which typically allows for actin polymerization [29].

Integrin dependent effects mediated by soluble CAMs are not without precedent. For example, in one report focused on L1CAM shedding, soluble L1 ectodomain stimulated cell migration in an integrin dependent manner. The authors suggested that L1 shedding might influence cell migration in both an autocrine and paracrine manner [46].

IV. MMPS, CAM CLEAVAGE, AND MICROGLIAL ACTIVATION

Classical microglial activation, in which release of specific proinflammatory molecules is increased [47, 48], occurs in the setting of HAND with substance abuse and likely plays a role in associated neuropathology. In several studies, HIV infection has been associated with an increase in the products of classically activated microglia, and it has been suggested that microglial activation may be an important mediator of CNS injury in the setting of both simian immunodeficiency virus encephalitis (SIVE) and HAND [49–51]. Minocycline has been linked to reduced microglial activation and disease severity with SIV infection, and SIV disease progresses more rapidly in animals showing an increase in markers of microglial activation. Increased activation of microglial cells has also been linked to the severity of HAND. Markers of macrophage/microglial cell activation are increased in association with disease severity, and apoptotic neurons are closely associated with markers of microglial activation [52]. Microglial activation also occurs with MA. Dopamine release and dopamine quinones in particular may be contributory [53]. Increased levels of reactive oxygen and nitrogen species may also play a role in that MA toxicity is reduced in copper/zinc superoxide dismutase transgenic mice [54].

Varied neuronal subpopulations may be vulnerable to neurotoxicity occurring as a result of microglial activation. These include glutamatergic neurons of the cortex and dopaminergic neurons of the substantia nigra. In animal models of Parkinson’s disease, microglial activation has been well linked to dopaminergic cell loss [55, 56]. Bacterial lipopolysaccharide, the prototypical inducer of strong microglial activation, also stimulates dopaminergic cell loss [57].

Studies suggest that MMP activity plays an important role in microglial activation. This claim is supported by studies in varied disease models. In a model of Parkinson’s disease, microglial activation, superoxide production and dopaminergic neuronal cell death were largely attenuated in MMP-3 null mice [55]. It has also been shown that inhibition of MMP-3 and -9 decreases LPS associated microglial activation [58]. Additional evidence for the importance of MMPs to microglial activation comes from studies with minocycline. In studies related to varied diseases and disease models, including Parkinson’s disease and multiple sclerosis, minocycline has shown neuroprotective effects [59]. Minocycline and other tetracycline derivatives can reduce MMP activity by at least two mechanisms including inhibition of microglial activation [49], which would reduce expression of microglial-derived MMPs, and direct inhibition of preformed MMPs via chelation of the active site zinc atom [60].

While the mechanisms by which MMPs activate microglia have not been well characterized, their ability to generate soluble integrin binding ligands may be particularly important. Microglia abundantly express varied integrins including the β₂ integrins αLβ₂ (LFA-1) and αMβ₂ (Mac-1). Microglia also express β₁ integrins α₄β₁ (VLA-4), α₅β₁, and α₆β₁ as well as the αv integrin αvβ₆. Expression of α₁, α₂, α_x, β₄, and β₇ has not been detected [61]. In terms of integrins that play a role in microglial activation, β₂ containing integrins may be particularly important. A recent study showed that Mac-1 [α_Mβ₂] was critical to the microglial activation of in a Parkinson’s Disease model. Activation of Mac-1 was associated with neurotoxin release and inhibition of Mac-1 signaling prevented neurodegeneration [62]. In a related study, Mac-1 was essential to β-amyloid induced microglial activation, production of superoxide, and neurotoxicity [63]. LFA-1 [α_Lβ₂] also plays an important role in microglial activation. Microglial cells are the principal resident cells of the CNS to express LFA-1 [64]. Engagement of LFA-1 has been linked to microglial spreading and to transcription of AP-1 responsive genes [64], which may lead to an activated phenotype. Similar to β₂ containing integrins, those containing β₁ might also be important to microglial activation. β₁ ligands can stimulate microglial cell activation and expression of pro-MMP-9 [65].

Shed CAMs interact with β integrins expressed on microglia. For example, VCAM has been shown to engage α₄β₁/VLA4, ICAM-1 has been shown to interact with LFA-1 and Mac-1, and ICAM-5 to interact with both both LFA-1 and β₁ [29, 44, 45]. As stated earlier, in recently published work, we have observed that soluble ICAM-5 stimulates integrin dependent phosphorylation of cofilin, an event which allows for actin polymerization. Actin polymerization occurs with microglial activation and, as will be discussed, is also thought to occur with changes in the structure of postsynaptic structures known as dendritic spines.

The expression of microglial integrins may be upregulated by inflammatory stimuli known to be elevated with HAND. Thus, an increase in MMP generated integrin binding ligands might assume greater significance in the setting of HAND. TNF-α can enhance surface expression of α₄ and LFA-1, while both IL-1 and TNF-α can increase expression of Mac-1 [61]. Moreover, integrin affinity for ligands is regulated by “inside out signaling” in which select stimuli influence the intracellular milieu to increase an integrin’s ability to engage its ligand [45]. Like expression, integrin affinity may also be increased in the setting of inflammation [66]. For a schematic depicting CAM cleavage and subsequent effects on microglia, please see Fig. (1).

Fig. (1). MMPs generate integrin binding ligands to activate microglia.

Schematic diagram showing: 1. Increased MMP activity in the setting of HAND with substance abuse; 2. MMP- dependent shedding of CAM ectodomains. Note that following CAM shedding, remaining C-terminal fragments may be further processed by intramembranous proteolysis and degradation; 3 and 4. CAM ectodomain engagement of microglial integrins to stimulate classical activation with 5. increased release of potentially neurotoxic molecules.

V. MMPS, CAM CLEAVAGE, AND SYNAPTIC FUNCTION

While studies of MMPs in the CNS have generally focused on CNS inflammation and injury, recent evidence suggests that MMPs play a critical role in normal CNS physiology [67].

Neuronal activity has been linked to increased MMP release [37, 68–70] and in recent experiments, we have detected MMP dependent shedding of a neuronal substrate within 5 minutes of N-methyl-D-aspartic acid (NMDA) treatment [70]. Recent studies suggest that MMPs exist in perisynaptic vesicular stores [12] and that vesicular MMPs from fibroblasts may be soluble NSF attachment protein receptor [SNARE] dependent [13]. As suggested above, if neuronal release is also SNARE dependent, it may occur with select stimuli such as neurotransmitters that increase intracellular calcium.

While a relatively new area of investigation, MMP activity has been shown to influence dendritic spine morphology as well as learning and memory [71–75]. An increase in the size of spines, allowing for increased post synaptic neurotransmitter receptor abundance, occurs in many studies of long-term potentiation (LTP). Though effects may be influenced by MMP dose/duration, and the developmental stage of neurons, these enzymes have the potential to increase spine size and hippocampal dependent memory. For example, MMP inhibition has been shown to reduce multiple forms of hippocampal CA1 plasticity [76], and MMP-9 activity has been implicated in the maintenance of late LTP [73]. In addition, MMP antisense constructs have been shown to prevent acquisition in the Morris water maze test [74], and at least one MMP can increase dendritic spine size in a manner that is temporally coordinated with LTP [72].

MMPs may also play a role in the maladaptive sort of learning that underlies addiction. That MA associated changes in MMPs contribute to addiction is supported by behavioral studies. For example, MA-induced behavioral sensitization and conditioned place preference, a measure of the rewarding effect of a drug, is reduced in mice lacking MMP-2 or MMP-9 [77]. Of interest, MMPs also play a critical role in cocaine associated conditioned place preference [78–80].

Though the mechanisms by which MMPs affect synaptic structure and function are not well understood, it is tempting to speculate that CAM cleavage plays a role. For example, ICAM-5 is highly expressed on thin dendritic spines and MMP dependent ICAM-5 shedding has been linked to spine maturation. Shedding of specific CAMs may allow for spine expansion. A non-mutually exclusive possibility that we have investigated is the possibility that the shed ectodomain can bind to unengaged postsynaptic integrins to stimulate dendritic actin polymerization and spine expansion. Integrin signaling is critical during developmental changes in spine morphology [81], and multiple forms of learning associated plasticity, including that mediated by MMPs, are thought to be integrin dependent [73, 76].

While relatively low levels of MMPs and/or MMPs released in a physiologically localized manner may enhance learning and memory, it should be noted that in HAND with superimposed substance abuse, levels of these enzymes may be pathologically elevated. This could stimulate excessive cleavage of CAMs that otherwise maintain synaptic structure and neuronal survival. It could also lead to a situation in which processes governed by physiologically appropriate MMP release go awry.

Several publications support the possibility that high levels of MMPs may be toxic. While this can occur by indirect mechanisms, in which BBB integrity is first disrupted, relatively direct mechanisms occur as well. For example, exogenous MMP-1 has been linked to neuronal death in dissociated and organotypic neuronal cultures. MMP-9 has also been shown to be directly neurotoxic. Moreover, high levels of exogenous MMP-7, which has a broad substrate range and cleaves varied synaptic CAMs, can stimulate synaptic injury as detected by changes including a reduction in the area of the post synaptic density [82]. Consistent with this, synaptic injury in the setting of brain trauma has been observed with increased MMP levels, and trauma associated reduction in synaptophysin immunoreactivity were diminished by MMP inhibition [11]. Synaptic injury occurs in HIV with substance abuse [83, 84], and it is tempting to speculate that MMPs, and the cleavage of specific CAMs in particular, play a role. This could occur both through the activation of microglia and possibly, by excessive proteolysis of neuronal and synaptic CAMs. Cell death has indeed been observed as a result of a disruption in adhesion, a process known as anoikis [85].

VI. CONCLUSIONS

MMP levels may be elevated with HIV infection and MA abuse. CAMs represent an important class of MMP substrates, which are ideally positioned to be processed by cell surface and secreted MMPs. Since the integrity of CAMs is critical to synaptic stability, excess cleavage of synaptic adhesion molecules may play a role in HIV/MA associated synaptic and neuronal injury. In addition, shed CAM fragments may interact with microglial integrins to stimulate classical activation of this cell type. The products of classically activated microglia can in turn compound neuronal injury. Future studies are warranted to determine whether MMP inhibitors or specific microglial integrin antagonists might be beneficial for the treatment of CNS inflammation occurring with HIV and MA.