In this study, we show that MMP9 expression is increased in cases with CAA-ICH compared to CAA cases without ICH. Furthermore, we show that there is more extensive TIMP3 staining in CAA cases versus controls and that a subset of CAA-ICH cases has a remarkable loss of TIMP3 expression.
MMP9 has been implicated in the development of ICH and BBB disruption [18, 28, 33,34,35, 39, 46]. In Tg2576 mice, a model for CAA and AD, microhaemorrhages were associated with MMP9 expression . Administration of recombinant MMP9 to the surface of mouse brains (through craniotomies) resulted in the occurrence of lobar haemorrhages, and Tg2576 mice with severe CAA were more susceptible to this procedure compared to wild-type mice . Human studies revealed increased levels of MMP9 in haemorrhagic areas compared to the contralateral hemisphere of CAA cases that suffered from ICH . Furthermore, Prussian blue positive products, indicative of cerebral microbleeds, were observed in the proximity of a cluster of vessels showing MMP9 immunoreactivity in CAA cases . However, all these findings were either of a qualitative, rather than of a quantitative nature , deduced from animal research [19, 55], or focussed on the brain region directly affected by ICH .
In our study, we aimed to quantitatively assess the expression of MMP9 in the human cerebrovasculature, without restriction to the immediate proximity of the haemorrhagic site (Table 1). Also, for the first time, we compared cases affected by CAA with and without ICH. We show here that increased vascular MMP9 levels are observed in CAA-ICH cases compared to CAA-NH, driven by a shift from partial MMP9 staining towards full MMP9 staining. Since we studied cerebral vessels distant from the site of ICH, our studies suggest that in CAA-ICH there is a global increased expression of MMP9, which may be mechanistically linked to the ICH. In correspondence with previous findings , Aβ-negative cortical vessels of CAA cases showed only minimal MMP9-immunoreactity. We have confirmed previous findings on a correlation between MMP9 staining and CAA grade  and on colocalization of MMP9 and Aβ [13, 19], both of which are a logical consequence of the capacity of MMP9 to degrade both soluble and aggregated Aβ [4, 15, 52, 53], and its induced expression in response to Aβ [9, 19].
Our findings on TIMP3 expression are in line with previous publications that reported that TIMP3 expression was more pronounced in CAA patients compared to controls, that TIMP3 colocalized with Aβ in leptomeningeal cerebrovascular arteries , and that TIMP3 protein levels were increased in brains of AD patients and a mouse model of AD . In addition, we showed that TIMP3 expression is not restricted to leptomeningeal arteries , but also detected in cortical arteries. We also demonstrate for the first time that a subset of CAA-ICH cases has a remarkably low expression level of TIMP3. Again, as we studied cerebral vessels not in the immediate vicinity of the site of ICH, our observations suggest that TIMP3 expression may be globally decreased in (a subset of) CAA-ICH cases, which may be mechanistically linked to the ICH. Not all CAA-ICH cases had decreased TIMP3 expression, indicating that other factors and pathways besides TIMP3 expression influence ICH development.
As TIMP3 has been shown to inhibit MMP9 activity [6, 8], we speculate that TIMP3 expression increases in response to elevated levels of MMP9 and possibly other MMPs in CAA patients, but that this negative feedback mechanism seems to fail in a subset of CAA-ICH cases, resulting in decreased inhibition of MMP9 and therefore increased risk of ICH development. The altered MMP9:TIMP3 ratio that we observed in CAA-ICH cases points towards a disbalance of these proteins. In addition, the positive correlation between the numbers of MMP9- and TIMP3-stained cortical vessels in CAA-NH cases supports the hypothesis of a feedback mechanisms in Aβ-affected vessels. In contrast, the absence of such a correlation in a subset of CAA-ICH cases may indicate that the feedback mechanism is failing in these patients. However, further, mechanistic studies are needed to confirm these hypotheses. Interestingly, increased TIMP3 levels may also contribute to an increased risk of CAA-related ICH, through activation of another pathway, as cellular overexpression of TIMP3 has been shown to redirect amyloid precursor protein (APP) processing towards the amyloidogenic pathway, through inhibition of ADAM10, a metalloproteinase that serves as an α-secretase . By reducing α-cleavage of APP and increasing β-cleavage, Aβ levels may increase as a consequence of TIMP3 overexpression . In the absence of ADAM10 expression data it is not possible to draw conclusions on the potential role of this latter mechanism.
The differential expression patterns of MMP9 and TIMP3 in CAA-ICH cases compared to CAA-NH cases were more pronounced in leptomeningeal vessels compared to cortical vessels. This may, in part, be due to the earlier and more severe accumulation of Aβ in leptomeningeal vessels as compared with cortical vessels [1, 38]. As we showed that both MMP9 and TIMP3 are strongly associated with Aβ, the observed differences in MMP9 and TIMP3 expression may therefore be more pronounced and easier detectable in leptomeningeal vessels. Furthermore, it is possible that altered MMP9 and TIMP3 levels in leptomeningeal vessels may disturb vascular functioning and blood flow and lead to haemorrhages downstream in cortical vessels, which may be in line with a recent observation that vessels do not rupture at the site of Aβ deposition [40, 41], but rather downstream or upstream, possibly mediated by impaired autoregulation. Furthermore, one could speculate that decreased integrity of leptomeningeal vessel walls may contribute to the development of cortical superficial siderosis, which is the deposition of blood-breakdown products in the subarachnoid space and strongly linked to CAA [7, 20].
Small haemorrhagic lesions are frequently observed in CAA patients, and we cannot rule out that such lesions may have influenced results. However, we assessed the presence of small haemorrhagic lesions in the brain tissue by Perls Prussian Blue iron staining. In several CAA-ICH cases, microbleeds were detected (supplementary Table 2). There was no appreciable staining of MMP9 or TIMP3 in the close proximity of microbleeds (Additional file 5), suggesting that such lesions may not or, only to a minor extent, have contributed to the observed differences in MMP9 or TIMP3 expression. Furthermore, even in the absence of (micro) hemorrhages, CAA-affected vessels may be leaky and permeable to plasma proteins , potentially resulting in MMP9 upregulation. However, fibrinogen immunostaining, as a proxy of BBB permeability, did not differ between CAA-NH and CAA-ICH cases and did not correlate with MMP9 staining (Additional file 6), making it unlikely that the upregulation of MMP9 is a result of increased vessel permeability.
Taken together, MMP9 and TIMP3 are directly associated with the presence of CAA. In addition, levels of these proteins are altered in CAA-ICH cases compared to CAA-NH cases. As altered levels of these proteins are directly related to an increased risk of ICH, MMP9 and TIMP3 pathways may have potential as therapeutic targets to prevent ICH in CAA patients. Noteworthy is our observation that MMP9 and TIMP3 are expressed at the site of vascular Aβ accumulation, but not in parenchymal Aβ accumulation (plaques). Of note, although we specifically assessed MMP9 and TIMP3, other members of the MMP and TIMP families may play a role in CAA-related ICH, such as MMP2, which has been previously associated with CAA-related ICH .
Our study has several limitations. First, our CAA cohort consists of a heterogenous patient group, with, in addition to moderate to severe CAA, varying degrees of AD pathology. Second, it is possible that not all cases of ICH were only due to CAA, and that other age-related pathological mechanisms were involved. Possibly, different aetiologies of ICH may explain the different TIMP3 expression patterns observed in our CAA-ICH cohort. Furthermore, as tissue of CAA-ICH cases is relatively scarce, we have included tissue from different brain banks, and therefore, we cannot rule out that differences in post-mortem interval and tissue treatment protocols (e.g. formalin exposure) may have affected our results. Finally, the number of cases included in our study was relatively small, and the results of this pilot study need to be validated in a larger cohort.
A strong point of our study is our unique cohort, which enabled us a direct comparison of CAA-ICH cases with CAA-NH cases. Previous observations on increased MMP9 levels in CAA-ICH are based on protein expression in proximity of the haemorrhagic area, compared to expression levels in the contralateral hemisphere , which may reflect post-ICH inflammation rather than a pathophysiological mechanism of ICH [35, 42, 48]. Another strong point of our study is the analysis of brain regions distant from the haemorrhagic site. In all cases, we studied the occipital cortex, whereas the haemorrhage usually had occurred in other locations, including the parietal and frontal cortices. Therefore, our data suggest that the observations of increased levels of MMP9 and decreased levels of TIMP3 may be a cause rather than a consequence of ICH, and that post-ICH inflammatory processes possibly only made a minor contribution to the observed differences, as we observed globally changed protein expression levels in the brains of CAA-ICH cases. However, we cannot rule out the possibility that MMP-9 levels in CAA-ICH cases increased post-ICH, especially in case of long time spans between ICH and death. However, previous reports on patients with haemorrhagic stroke did not detect an increase of MMP9-positive vessels in contralateral brain sections, in contrast to perihematomal tissue [13, 35].
In conclusion, we provide evidence that increased cerebrovascular levels of MMP9 and decreased levels of TIMP3 are associated with CAA-related ICH. Future studies are needed to validate these findings in larger data sets, and to determine the mechanistic pathways leading to the altered expression levels of MMP9 and TIMP3.