Microhemorrhages are a significant consequence of cerebral amyloid angiopathy (CAA) and based on MRI detection, are known to occur in as many as 40% of AD cases [32
]. As imaging techniques become more sensitive, this percentage may become even higher. In addition, microhemorrhages are exacerbated by anti-Aβ immunotherapy in mouse and human studies [8
]. Another vascular adverse event of immunotherapy is vasogenic edema, which has been found in the ongoing immunotherapy trial of bapineuzumab [5
]. Vasogenic edema involves disruption of the blood-brain barrier and may therefore share a common mechanism of action with microhemorrhages. While a number of pathophysiological factors may contribute to microhemorrhage and vasogenic edema, proteolytic destabilization of the neurovascular unit may be an important feature. Matrix metalloproteinases are a family of proteinases known to degrade components of extracellular matrix, as well as other proteins including cytokines and pro/anti-angiogenic modulators. In particular, MMP2 and MMP9 are implicated in cerebrovascular dysfunction. Using mouse models of AD, we have examined changes in these MMP systems with immunotherapy in two separate studies, an active vaccination study and a passive immunization study, both of which have shown increased incidence of microhemorrhage.
MMP2 is also known as gelatinase A and is secreted as a pro-form, which requires cleavage by MT1-MMP (MMP14) for its activation [33
]. MT1-MMP is activated by furin and plasmin. In addition, plasmin has been shown to directly activate surface-bound pro-MMP2 [34
]. An endogenous inhibitor of MMP2, TIMP2, controls the activity of MMP2 post-translationally [35
]. Importantly, MMP2 has been shown to be involved in the early opening of the blood brain barrier following cerebral reperfusion [36
]. This MMP2-mediated process involves degradation of the tight junction protein, claudin-5 [37
In order to evaluate the activity of the MMP2 system in our mouse models of AD, we examined the gene and corresponding protein expression levels of furin, MT1-MMP, MMP2 and TIMP2 in mice that received active Aβ vaccination. We found that vaccination resulted in increased gene expression of furin, and MMP2 along with a concomitant decrease in TIMP2 expression, while protein expression followed the same pattern. In addition, we established that MMP2 activity was significantly increased by performing gelatin zymography on brain lysates. These observed changes are consistent with an increase in proteolytic activity that may degrade collagen or other extracellular matrix proteins that comprise the blood brain barrier, leading to leakage of cerebral vessels. To confirm that our findings were relevant to immunotherapy, we obtained frozen hippocampi from APPSw mice that were passively immunized and showed significant microhemorrhage incidence [11
]. Indeed we found similar trends in passively immunized mice as compared to the actively vaccinated mice. Following one month of immunization MMP2 was increased and for the following two and three months only the MMP2 remained significantly elevated. TIMP2 levels, however, transiently and slightly increased at 1 month then remained at a reduced level. While it may seem that the active vaccination had a lesser effect on the MMP2 system than the passive immunization, it is important that we not draw conclusions from this since these are different strains of mice (APPSw/NOS2-/- vs. APPSw) and different aged mice (16 vs. 22 months). All of these things could account for the apparent smaller effect size in the active vaccination study. Overall, the MMP2 system data suggest that increased activity of the MMP2 system may be involved in abnormal proteolytic activity at the neurovascular interface.
The second major proteolytic enzyme in the brain, MMP9 (also known as gelatinase B), has been heavily implicated in many types of CNS injury including stroke [38
], ischemia [35
] and trauma [39
]. The normal function of MMP9 is degradation of extracellular matrix to allow for cell migration and also degradation of basement membranes to allow for cell movement across the blood vessels. In addition, Aβ is a natural substrate of MMP9 indicating a role in Aβ homeostasis [40
]. MMP9 is secreted in a pro form, which is cleaved to the active MMP9 by other MMPs, primarily MMP3. Inflammatory cytokines such as IL-8 and TNFα [22
] are known to be involved in the regulation of the MMP9 system through the induction of both MMP3 and MMP9. Also, TNFα [41
] and IL-1β [42
] are examples of cytokines that are actually activated by MMP9. Similar to MMP2, an endogenous inhibitor of the proteolytic activity of MMP9 has been described, and is known as TIMP1. TIMP1 provides an additional level of regulation of proteolysis in the brain and can be regulated, in turn, by inflammatory factors and tissue redox balance [35
The MMP9 system was assessed by qRT-PCR, ELISA, zymography and immunohistochemistry in our Aβ vaccination study in APPSw/NOS2-/- mice. We found that MMP3 gene and protein expression are robustly increased, while MMP9 is increased to a lesser extent. TIMP1 levels were reduced at the gene expression level and remain unchanged at the protein level. A similar pattern was observed at 2 and 3 months in the passive immunization study. In addition, zymography revealed that MMP9 activity is significantly increased following Aβ vaccination in APPSw/NOS2-/- mice. Finally, MMP9 immunohistochemistry showed a dramatic elevation of expression in the endothelial cells of the cerebrovasculature in APPSw/NOS2-/- mice receiving Aβ vaccination compared to those mice control vaccination.
In vitro MMP2 and MMP9 have both been shown to proteolyze Aβ [44
]. Moreover, in vivo MMP9 is increased in association with amyloid plaques in transgenic mouse models of amyloid deposition [46
]. Additionally, treatment of APP/PS1 mice with FK506 lowered Aβ and increased MMP9 suggesting a relationship between elevated MMP9 and lower Aβ [47
]. However, in human AD despite MMP9 being overexpressed around amyloid deposits [48
], MMP2, MMP3 or MMP9 cannot be associated with plaque load in AD suggesting that they are not crucial for the regulation of plaque load [49
]. We believe that the upregulation of MMPs in our immunotherapy studies may play a small role in the clearance of Aβ due to immunotherapy but other clearance mechanisms are also known to play a significant role (reviewed in [4
]). Our data in BV2 cells indicates that MMPs, especially the MMP9 system can be upregulated through Fcγ receptor signaling mechanisms more efficiently than Aβ alone and this is likely the primary mechanism for MMP upregulation in our immunotherapy studies.
We have previously shown a robust inflammatory response to anti-Aβ immunotherapy, regardless of whether an active vaccination or passive immunization approach is used [50
]. To determine whether the inflammatory response to immunotherapy was, at least in part, responsible for the increased MMP2 and MMP9 activities, we used BV2 microglial cells to determine how immune complexes change the MMP expression. Treatment of BV2 microglial cells with immune complexes of anti-Aβ IgG/Aβ or anti-tau/tau significantly increased MMP3 and MMP9, the two MMPs most closely associated with inflammation. In contrast, we did not observe significant changes in MMP3 or MMP9 with IgG alone or Aβ alone, indicating an Fcγ-receptor mediated activation.