Matrix Metalloproteinase in Blood-Brain Barrier Breakdown in Dementia

MMP overview

MMPs are a family of 23 proteases in humans(24 in mice) that degrade the extracellular matrix along with other substrates such as contractile proteins, tight junction proteins, and pro-forms of signaling molecules [1]. Based on substrate specificity, MMPs are separated into five classes: collagenases, stromelysins, a heterogeneous group, membrane anchored MMPs, and finally, the gelatinases. The gelatinase class, which are present in the brain, includes MMP2, which is also known as Gelatinase A, and MMP9, also known as Gelatinase B. The gelatinase class of MMPs can degrade gelatin, cytokines [1], and even amyloid-β (Aβ) [2, 3], but they also have roles in axonal growth, synaptic plasticity, and vascularization [4–8]. Due to their variety of substrates, it is important to maintain control over MMP2 and MMP9 activity; therefore, they are endogenously regulated at several different levels. Similar to other MMPs, gene expression of MMP2 and MMP9 is regulated by cytokines, growth factors, and other proteins. The pro-inflammatory cytokines tumor necrosis factor α (TNFα) and interleukin 1β (IL1β) can stimulate transcription of MMP9 [9, 10] while activator protein 2, specificity protein 1, and polyomavirus enhancer-A binding protein 3 can stimulate transcription of MMP2 [11]. At the protein level, MMP2 is constitutively expressed in the brain while MMP9 expression is induced by neuroinflammation. However, both are secreted as zymogens that require cleavage via other MMPs. For MMP2 activation, a complex of MMP14, which is also known as membrane type 1 MMP (MT1-MMP), tissue inhibitor of metalloproteinase 2 (TIMP2) and proMMP2 must be formed. MMP14 is one of six enzymes that include a C terminal hydrophobic transmembrane domain that anchors the enzyme to the plasma membrane and thus restricts its activity to the cell surface [12, 13]; this then constrains the action of MMP2 as well [14]. Activation of proMMP9 requires cleavage via MMP3 which can also be activated via neuroinflammation [1]. Finally, the endogenous inhibitor TIMP1 binds and inhibits MMP9 [1]. Active MMP2 can be inhibited by TIMP2, even though TIMP2 is necessary, along with MMP14, for activation of proMMP2 [1]. Therefore, TIMP2 participates in activation of proMMP2 and inhibition of active MMP2. Recent studies have also shown that nitric oxide (NO) can regulate MMP9 activity and the activity of TIMP1 [15]. The MMP2 and MMP9 systems, which are summarized in Fig. 1 are associated with tight junction breakdown leading to BBB leakage [16].

Stroke

Cognitive impairment is common after stroke and can affect different cognitive domains, although attention and executive functions are the most commonly impacted [17]. Impairment in these areas is usually seen 3 months after stroke and the prevalence of cognitive impairment after stroke varies from 11.6% to 56.3% in various hospital-based studies [18–22]. Stroke also doubles the risk of dementia and is a major contributor to vascular contributions to cognitive impairment and dementia. While most studies focus on ischemic stroke, one study has shown that 60% of patients were cognitively impaired 3 months after a hemorrhagic stroke and had reduced spatial ability and executive function and a decline in Mini-Mental State Examination scores between baseline and at the first and second year follow-ups [23].

Cerebral microhemorrhages are also common after ischemic and hemorrhagic stroke with a prevalence of 18% to 71% in ischemic stroke and 47% to 80% in intracerebral hemorrhage (ICH) [24–26]. Microhemorrhages have been linked to a decline in executive function as well as memory, visuospatial impairment, and language. The Age, Gene/Environment Susceptibility-Reyjkavik study showed that multiple microhemorrhages were associated with worse processing speed and executive function [27]. The Rotterdam Scan Study showed that at least five microhemorrhages were associated with worse cognitive performances in all domains except memory [28]. Therefore, this opening of the BBB resulting in microhemorrhages can contribute to cognitive decline after stroke and the gelatinases have been heavily implicated in this breakdown.

The gelatinases function in neurovascular remodeling and microvascular recanalization and are also needed for modulating interactions between cells during development and tissue remodeling [29–32]. However, unregulated MMP activity can lead to an increase in BBB permeability, which results in vasogenic edema and hemorrhage, and contributes directly to neuronal injury, apoptosis, and brain damage after acute cerebral ischemia [33–36]. This disruption in the BBB that occurs early after an ischemic stroke is central to hemorrhagic transformation.

After a stroke occurs, the BBB around the ischemic area is weakened and bleeding into the brain can occur. Termed hemorrhagic transformation, this opening of the BBB occurs in 10% to 40% of patients with ischemic stroke [37, 38] and greatly increases the morbidity and mortality after stroke [39, 40]. The only FDA approved therapy for stroke, tissue plasminogen activator (tPA), increases the occurrence of hemorrhagic transformation by 10-fold when administered outside the narrow 3-h window when it is effective [41, 42]. Both MMP2 and MMP9 have been implicated in the hemorrhagic transformation that occurs after stroke and during delayed tPA administration.

While tPA works to restore blood flow to the ischemic area by breaking down fibrin based clots, it can activate and release MMP2, MMP9, and MMP3 [41, 43–45]. By binding to the protease activated receptor 1 (PAR1) and activating the NFκB pathway, tPA increases expression of MMP9 [46]. By binding to platelet derived growth factor receptor alpha (PDGFRα) on astrocyte end feet and increasing PDGF-CC, tPA also increases MMP2 expression and BBB permeability [47]. The lipoprotein receptor protein receptor on endothelial cells is another target of tPA and binding of this receptor can increase expression of MMP3 and MMP9 [44, 48]. Finally, neutrophil degranulation leading to the release of stored MMP9 into the blood is increased by tPA administration [49]. Thus, MMP9 may act as a blood marker for hemorrhagic transformation.

Hemorrhagic transformation after ischemic stroke can be separated into two phases: early phase (18–24 h after stroke onset) and late phase (24 h after stroke onset) [50]. During the early phase, leukocytes infiltrate the brain and are considered to be the main source of MMP9. Ischemic stroke quickly initiates an immune response causing leukocytes to bind to endothelial cells in the vasculature [51]. In humans, MMP9 mRNA levels in peripheral leukocytes are increased as soon as 3–5 h after stroke and activity of MMP9 peaks at 6–8 h after stroke [51–53]. Plasma levels of MMP9 have been shown to be predictive of hemorrhagic transformation and correlate with BBB injury [54–57]. MMP9 and MMP2 are increased in plasma 3–8 h after stroke in rats and mice as well [58, 59]. During the early phase of hemorrhagic transformation, the increase in plasma MMP9 can breakdown the BBB on the luminal side of the vessel by attacking tight junctions or can be transported across the endothelial cell and disrupt the neurovascular unit.

The brain is also an important source of MMP2 in the early phase of hemorrhagic transformation. Primarily derived from astrocytes and endothelial cells, MMP2 is increased in the post stroke brain within 1–3 h in rats and mice and remains elevated for several days [60–62]. MMP14 has also been shown to be increased after stroke, which could subsequently increase MMP2 activity [63, 64]. This rise in MMP2 correlates with degradation of claudin-5 and occludin leading to BBB breakdown in rodent models of stroke [61, 65–67]. Direct injection of MMP2 into the brain results in breakdown of the BBB and hemorrhage as well [66]. Inhibition of MMP2 with BB-1101 blocks MMP2 activity and results in a reduction in edema and BBB permeability at 24 h after stroke [65]. This effect did not continue 48 h after stroke though. In MMP2-deficient mice, there was a reduced rate of hemorrhagic transformation, smaller hemorrhage size, and improved neurological function after stroke [62]. However, other studies have shown that MMP2 levels were not significantly increased in animals with hemorrhagic transformation compared to those without and MMP2 knockout mice failed to reduce infarct size[68, 69]. Further exploration into the role of MMP2 during hemorrhagic transformation is needed.

In the late phase of hemorrhagic transformation, brain cells are the major source of MMPs. MMP9 has been shown to be expressed by astrocytes, neurons, microglia, and endothelial cells and MMP9 is localized to these immunoreactive cells during the late phase of hemorrhagic transformation [8, 70]. This suggests that leukocytes are no longer the major source of MMP9 at 24 h after stroke onset. In fact, Maier et al. showed that the major source of MMP9 at 24 and 72 h were brain cells rather than neutrophils [71]. In rats, when neutrophils are depleted, the brain levels of MMP9 and MMP2 are unchanged compared to controls at 24 h [69]. MMP3 was shown to be increased within 24 h after stroke onset [72], and MMP3 knockout mice were shown to have a reduced rate of hemorrhagic transformation in a MCAO model with tPA treatment [73]. This increase in MMP3 can increase the activity of MMP9 via cleavage of proMMP9, contributing to hemorrhagic transformation.

Due to the increase in MMP activity during delayed tPA administration and hemorrhagic transformation, the use of MMP inhibitors to reduce the risk of BBB breakdown has been studied. Mishiro et al. conducted a study where mice were subjected to 6-h filamental middle cerebral artery occlusion and treated with vehicle, tPA alone, or tPA plus GM6001, a broad spectrum MMP inhibitor [74]. In vivo studies show that inhibition of MMPs after delayed tPA treatment resulted in reduced hemorrhage volumes, a reduction in MMP9 activity, and an increase in occludin and zona occludens 1 (ZO-1) proteins compared to mice treated with tPA alone [74]. In vitro studies with GM6001 treatment after tPA treatment with hypoxia and reoxygenation resulted in reduced cell damage and an increase in transendothelial electrical resistance due to a reduction in degradation of occludin and ZO-1 [74]. Similar results were shown by Chen et al. where GM6001 treatment significantly reduced BBB breakdown and ameliorated brain edema by preventing the decrease in occludin and ZO-1 [75]. In a model of permanent focal cerebral ischemia, treatment with SB-3CT, a gelatinase inhibitor, led to a decrease in MMP9 activity, decreased basement membrane laminin degradation, prevented pericyte lumen from contraction, and protected pericytes and endothelial cells [76]. Repeated SB-3CT treatment can counteract degradation of neuronal laminin, protect neurons from ischemic cell death and ameliorate neurobehavioral outcomes after embolic MCAO in mice [76]. SB-3CT also abolishes oxygen glucose deprivation induced reduction of occludin as well as decreases Evans blue extravasation and apoptotic cell death after subarachnoidhemorrhage [61, 78]. A recent clinical trial indicated that minocycline, which acts an anti-inflammatory, can lower plasma MMP9 levels, even at 72 h post stroke, and improve neurological outcomes in acute ischemic stroke patients treated with tPA [79]. However, it is important to note that MMPs have also been implicated in the repair functions after cerebral ischemia such as neuroblast migration and neuronal plasticity; therefore, MMP inhibition may only be beneficial during a short window after stroke [8, 80]. Following injury, blood vessels are dependent on the plasminogen-activator system and on MMPs for their regeneration [81], and Lo et al. showed that delayed inhibition of MMPs by broad spectrum inhibitors is detrimental during the ischemic recovery phase seven days after cerebral ischemia, hindering brain repair in mice and attenuating neurovascular regeneration in the penumbra [8]. Further investigation into both long-term detrimental effects and the short term beneficial effects of MMP inhibitors is required to determine therapeutic treatments and windows.

While the role of MMPs in hemorrhagic transformation after ischemic stroke has been studied in depth, their role in hemorrhagic stroke or ICH is less clear. Human studies have revealed elevated levels of MMP9 in peripheral blood after intracranial hemorrhage and were associated with worse neurological outcomes [82]. Another study showed that while MMP2 levels remained stable after hemorrhagic stroke, MMP9 levels were increased and could possibly have contributed to cerebral edema after stroke [83]. However, MMP9 is more commonly associated with CAA hemorrhage.

When the BBB is compromised during ICH, the essential component to the coagulation cascade, thrombin, enters the brain [84]. The increased levels of thrombin after ICH are associated with the development of cerebral edema and injection of thrombin into the basal ganglia causes BBB leakage and cerebral edema [84–87]. Protease activated receptors (PARs) also mediate the development of edema after ICH and are one of the receptors for thrombin [88]. Pericytes treated with thrombin show increased expression of MMP9 and the use of a PAR1 antagonist leads to a decrease in MMP9 expression after thrombin treatment [89]. This suggests that thrombin acts through PAR1 to induce MMP9 expression in pericytes, thus leading to further BBB breakdown.

After an ICH, the secondary brain injury that arises from tissue reaction to blood products results in edema which is a crucial factor when treating ICH. Dexamethasone has been shown to treat edema after spinal cord injury and for brain tumors. Yang et al. showed that rats treated with dexamethasone after ICH had a reduction in edema and MMP9 as well as normal levels of IκB and reduced levels of NFκB [90]. This suggests that dexamethasone blocks NFκB signaling to reduce MMP9 levels and reduce edema, providing a possible therapy for edema after ICH.

CAA and AD

AD is the most common form of dementia and is characterized by Aβ plaques and neurofibrillary tangles consisting of hyperphosphorylated tau. Of the patients with AD, almost 95% have CAA or the accumulation of Aβ around arteries in the cerebral cortex [91]. The main criterion for clinical diagnosis of CAA is the presence of cortical cerebral brain hemorrhage which causes and/or contributes to neurodegeneration and dementia [92]. It has been shown that patients with AD and multiple microhemorrhages had more severe cognitive impairment compared to AD patients without microhemorrhages [93]. MMP2 and MMP9 have been widely associated with BBB disruption and microhemorrhages in CAA and AD.

Hernandez-Guillamon et al looked at MMP2 and MMP9 expression in CAA patients [94]. While plasma MMP2 and MMP9 levels did not discriminate between CAA patients and controls, MMP9 and MMP2 levels were higher in the hemorrhagic areas of CAA brains. proMMP9 and proMMP2 were significantly higher in hemorrhagic areas of patients who died from a fatal hemorrhage. MMP9 was mainly found in brain vessels and neurons in the area around the hemorrhage and the contralateral side while MMP2 was mainly located in medium sized brain vessels and capillaries and was more pronounced around the hemorrhagic area. MMP2 was also expressed in reactive astrocytes surrounding Aβ affected vessels and its expression was clearly associated with the Aβ load in the vessel in grade 1 and 2 CAA. Briefly, grade 1 CAA is given when some amyloid deposits are found in otherwise normal vessels, grade 2 CAA is used when the vessel media is replaced by amyloid, grade 3 CAA occurs when cracking of the vessel wall results in “vessel-within-vessel” appearance and affects at least 50% of the circumference of the vessel, and finally, grade 4 CAA is given when scarring of the wall is observed as well as fibrinoid necrosis and traces of intermingled amyloid deposits [95]. In grade 3 CAA lesions, Aβ reactivity was very frequent and was found to have MMP2 expression in astrocytes surrounding the lesions. Unlike MMP2, MMP9 was found in some isolated macrophages around the grade 3 CAA lesions. There was some Aβ reactivity in grade 4 CAA lesions and strong gliosis was prominent with reactive astrocytes positive for MMP2. MMP9 expression was similar to grade 3 CAA. In mice, MMP2 was found in reactive astrocytes around Aβ plaques and near cerebral microvascular fibrillar amyloid deposits in Tg-SwDI mice [96]. MMP9 expression was also shown in CAA vessels in amyloid-β protein precursor (AβPPsw) transgenic mice [97]. This suggests that MMP2 and MMP9 play a role in BBB breakdown and hemorrhaging in CAA.

Recent evidence suggests that Aβ may contribute to the BBB leakage that is associated with CAA. Isolated rat brain microvessels that are exposed to increasing concentrations of Aβ₄₀ show significant decreases in claudin-1 and claudin-5 expression as well as increases in MMP2 and MMP9 expression [98]. Microvessels from 9-month-old transgenic AβPP mice (Tg2576 strain with the Swedish mutation KM670/671NL) also show increased permeability, decreased claudin-1 and claudin-5 as well as increases in MMP2 and MMP9. Cultured human brain endothelial cells treated with Aβ₄₀ caused ZO-1 to retreat from the plasma membrane which in turn caused a decrease in transendothelial electrical resistance [99]. Carrano et al. found decreased occludin, claudin-5, and ZO-1 in brain microvessels of postmortem CAA patient brain slices [100]. Along with changes in tight junction proteins, there are also studies that show changes in the expression of MMPs in the CAA brain and that these increased MMPs contribute to the BBB breakdown. Mouse endothelial cells and an AβPPsw mouse CAA model had Aβ mediated upregulation of MMP9 [101]. Soluble Aβ₄₀ generates reactive oxygen species and pro-inflammatory cytokines which can activate MMP9 and induce cerebral hemorrhage [102].

The prevailing hypothesis for AD progression is the amyloid hypothesis, which, briefly, states that the deposition of amyloid leads to hyperphosphorylation of tau which in turn leads to neurodegeneration [103]. Using this hypothesis to develop therapeutics, anti-Aβ immunotherapy has become the most promising approach for AD. Unfortunately, in clinical trials, vascular adverse events like microhemorrhages and vasogenic edema were reported with anti-Aβ treatment [104]. MMP2 and MMP9 have been implicated in the BBB breakdown that occurs after anti-Aβ immunotherapy. In AβPPSw/NOS2-/- mice that received an active Aβ vaccination for 4 months, there was a significant increase in gene expression of furin and MMP2 and a decrease in TIMP2 [105]. ELISA measurements confirmed these qPCR findings. There was also an increase in gene expression of MMP3 and a decrease in TIMP1 levels in the actively vaccinated transgenics. Protein measurements showed increased MMP3 and MMP9 as well. Within this study, gelatin zymography analysis also showed a significant increase in MMP2 and MMP9 activity in the transgenic mice treated with the active Aβ immunotherapy. Robust expression of MMP9 was also found to be associated with the vasculature. Similar results were found in passively immunized AβPPSw mice. In vitro studies using BV2 microglial cells showed that Aβ₄₂ fibrils increased MMP2 and MT1-MMP gene and protein levels [105]. An increase in MMP9 and MMP3 occurred when BV2 cells were treated with either an anti-Aβ IgG-Aβ or anti-tau IgG-tau immune complex, showing that MMP9 activation requires the Fcγ receptor [105].

It has long been known that AD induces a neuroinflammatory response, although the effect of this inflammation remains unknown. During AD, peripheral immune cells, such as lymphocytes, monocytes, and neutrophils, can cross the BBB and infiltrate the brain [106]. Aβ₂₅₋₃₅, a truncated form of the Aβ peptide, has been shown to activate invading neutrophils, the most abundant immune cell to infiltrate the brain, and stimulate the release of proMMP9 [107]. While this stimulation and release does not activate MMP9, there are several other proteases, such as MMP3, that can cleave and activate the released MMP9. This activation of neutrophils and release of proMMP9 indicates that MMP9 may play a role in the inflammatory response during AD.

While there have been several studies looking at the effect of Aβ on BBB breakdown and MMPs, there are even fewer studies looking at the effect of tau. One study has shown that rTg4510 mice, which normally show tau accumulation and neuronal loss at 2.5 months of age and gliosis at 4 months of age, develop BBB breakdown at 9 months of age [108]. While this disruption in BBB does not coincide with tau accumulation or neuroinflammation, it does appear along with perivascular tau suggesting that perivascular tau causes BBB disruption. Since the rTg4510 mice contain a tetracycline-controlled transactivator, doxycycline can be administered to reduce tau levels. Reduction of tau recovered the BBB defect seen at 9 months [108]. However the mechanism and the role of MMPs in this breakdown remains unclear.

Tau has also been shown to be a substrate for MMP3 and MMP9 [109]. While MMP3 cleavage of tau does not result in tau aggregation, MMP9 cleavage does result in tau oligomer formation. Frost et al.has previously demonstrated that extracellular pro-aggregatory tau can be incorporated into neurons and induce neurotoxic intracellular tau aggregation which spreads to other cells [110]. This suggests that cleavage of tau by MMP9 could contribute to neurofibrillary tangle formation.

AD is also characterized by a progressive downregulation of specific cellular markers within the cholinergic basal forebrain nucleus basalis neurons that provide the major cholinergic innervation to the entire cortical mantle [111]. The degeneration of these cholinergic neurons and reduction in cortical choline acetyltransferase activity has been associated with cognitive decline in AD [112]. The leading hypothesis for the vulnerability of these neurons may be attributed to the loss of trophic support in AD, specifically nerve growth factor (NGF) and its receptors [113]. It has been shown previously that the NGF receptors, TrkA and p75^NTR_, decline early in AD and correlate with cognitive dysfunction [114, 115].

proNGF has recently been discovered to play a key role in biological processes in neuronal activity along with mature NGF (mNGF). Degradation of NGF occurs via MMP9 and MMP9 has been found to be elevated in the cortex of patients with AD [116, 117]. This alteration in MMP9 may contribute to the vulnerability of neurons during AD. In postmortem AD brains, MMP9 is elevated in neurons, senile plaques, tangles and within the vascular wall [118]. MMP9 is also elevated in the plasma, hippocampus, and cerebral cortex of AD patients [117–120]. Bruno et al. demonstrated an increase in MMP9 activity during the progression of AD that correlates with cognitive impairment [121]. Since MMP9 plays a role in degrading mNGF in the CNS, it can be suggested that degradation of NGF via MMP9 may play a role in the cognitive decline that occurs in AD.

CONCLUSIONS

Over the years, MMP2 and MMP9 have been increasingly implicated in the BBB breakdown that contributes to cognitive decline after stroke and in CAA and AD. Both MMP2 and MMP9 have been shown to be increased during hemorrhagic transformation after stroke and during delayed tPA treatment. The gelatinases have also been shown to be increased around vessels with Aβ and hemorrhaging and are increased after anti-Aβ immunotherapy. While, the source and location of these MMPs in these diseases have been elucidated and some studies have shown a therapeutic benefit of MMP inhibitors after stroke, further studies are needed to determine both long term detrimental effects and the short term beneficial effects of these inhibitors as well as effects of MMP inhibition on microhemorrhage occurrence in CAA and as an adjunct therapy with anti-Aβ immunotherapy.