Early changes in extracellular matrix in Alzheimer's disease

Although changes in extracellular matrix (ECM) scaffold have been reported previously in Alzheimer's disease (AD) compared to normal ageing, it is not known how alterations in the numerous components of the perivascular ECM might occur at different stages of AD. This study therefore investigates potential changes in basement membrane‐associated ECM molecules in relation to increasing Braak stages.


Introduction
Alzheimer's disease (AD) is a chronic and progressive neurodegenerative disease, and is the most common form of dementia amongst the elderly population. It is characterized by memory loss and cognitive decline. Among the hallmarks of AD, the most well characterized are the presence of numerous extracellular (senile) plaques (SP) and neurofibrillary tangles (NFT) in neocortical and some subcortical brain regions [1]. The extracellular plaques consist of aggregates of amyloid b protein (Ab), which is a breakdown product of amyloid b precursor protein (APP) when cleaved by band c-secretases. NFT result from an aggregation of abnormally phosphorylated microtubule-associated tau proteins.
In the cerebral cortex, amyloid deposits and tau aggregation show a characteristic distribution pattern following a predictable sequence over the course of the disease [2,3]. It has previously been described in animal models that extracellular matrix (ECM) components can influence b amyloid fibrillogenesis, and thereby may contribute to AD progression [4,5]. Moreover, Ab seems to interact with a variety of proteoglycans and fibrous proteins [6][7][8][9][10][11]. Ab deposition, as well as neurodegeneration, is highly damaging to the integrity of the pre-existing ECM, resulting in a deterioration of physico-chemical properties, such as diffusion parameters of the extracellular micromilieu [12,13].
The ECM constitutes the essential physical scaffolding for cells, and acts as a biomechanical and biochemical support required for tissue rearrangement, differentiation and homeostasis [14]. The functional attributes of the ECM in the adult central nervous system (CNS) arise directly from their molecular composition, which is localized into three principal compartments: the neural interstitial matrix, the perineuronal nets and the basement membrane. The latter is a sheet-like layer that serves as a boundary between CNS parenchymal tissue and endothelial cells and is predominantly made up of collagen, fibronectin and proteoglycans such as perlecan [15]. Collagen type IV is the most abundant fibrous protein within the interstitial ECM. Collagen IV is essential for regulating cell adhesion and directing tissue development through its role in chemotaxis and cell migration, while also providing strength to the ECM [16]. Perlecan interacts with several other components of the ECM, and is an essential component of the vascular ECM involved in maintenance of the endothelial barrier function [17]. Fibronectin is directly involved in mediating cell attachment and function and, as such, in directing the (re-)organization of the interstitial ECM [14]. Another component of the basement membrane, the human platelet endothelial cell adhesion molecule (hPECAM), has a higher level of expression at the endothelial cell junction, is involved in cell adhesion, and acts as a signalling molecule [18] through interactions with the ECM [19].
ECM changes have been reported in both normal ageing and in AD, and this remodelling could mediate cell proliferation, differentiation and apoptosis [20,21]. The functional consequences of this remodelling may be dependent on the local specific organization of the ECM as established in animal models [22][23][24][25]. Any, or all, of these key ECM components might be affected during AD pathogenesis, though previous reports have described contradictory changes in ECM components in AD. For example, some reports found an increase in collagen IV in cerebral microvessels in AD and Parkinson's disease [26][27][28], while other reports have shown a decrease [29] or no change in intensity of staining [30] in cortical vessels in AD. As regards proteoglycans such as perlecan, these ECM components have been shown to be present in both amyloid plaques and NFT. They are thought to contribute to disease pathogenesis by accelerating the formation of Ab through preventing proteolytic degradation [10,31]. Total uronic acids, as basic components of the proteoglycan perlecan, have been reported to be increased in cerebral microvessels in AD [26]. Conversely, no significant difference was seen in perlecan mRNA levels in the hippocampus in AD when compared to age-matched controls, indicating that perlecan expression may remain stable in AD [32]. In the case of fibronectin, it has been previously reported that high molecular fibronectin forms appear more frequently and at higher amounts in plasma in AD than in age-matched controls [33], and that levels of plasma fibronectin are higher in AD than in healthy controls and individuals with mild cognitive impairment [34]. On the other hand, unchanged plasma fibronectin concentrations have also been described in AD [35].
To date, no studies have compared the expression of different ECM components within the same groups of control and AD patients, or have investigated changes in ECM components in relation to disease progression, as classified by Braak staging [2]. Therefore, the aim of the present study was to identify changes in ECM in healthy control subjects with low (Braak ≤2), intermediate (Braak >2 & ≤4) and high Braak stages (Braak >4). We have used immunohistochemistry to look for changes in three major ECM componentscollagen IV, perlecan, fibronectinas well as a cell adhesion and signalling molecule, hPECAMin the cerebral cortical grey matter of these individuals. We also assessed the severity of Ab plaque formation to test for potential correlations between plaque load and potential changes in collagen IV, perlecan, fibronectin and hPECAM. Finally, we assessed capillary density by immunostaining for the von Willebrand factor (vWF), a multimeric glycoprotein present in blood plasma and produced largely in endothelium and megakaryocytes, which has been widely used to demonstrate vessels in tissue [36].

Patients
Thirty cases (9 men and 21 women) were investigated, and post-mortem brain tissues were obtained from the Manchester Brain Bank through appropriate consenting procedures for the collection and use of human brain tissues (Manchester Brain Bank Generic Tissue Bank Ethical Approval number 09/H0906/52). Pathological diagnoses were made by an experienced neuropathologist (DM) as previously described [37,38], and in accordance with recent National Institute on Ageing -Alzheimer's Association guidelines for the neuropathological assessment of AD [39]. By such classification, only cases with Braak stages 5 and 6 were classed as definite AD. Cases with Braak Stages >2 & ≤4 were not diagnosed as definite AD, but these could be considered as cases of possible (incipient) AD, thereby representing cases with early changes of AD. Amyloid plaque load per se was not employed for classification of AD, although it is acknowledged that neuritic plaque scores do form the basis of Consortium to Establish a Registry for Alzheimer's Disease classification. Clearly, in terms of amyloid plaques alone, there was overlap in amyloid load between cases at Braak stage >2 & ≤4 and those at Braak stages 5 and 6. Subjects were therefore split in three groups based on their Braak stages [2]: controls (Braak ≤2, 12 subjects), subclinical AD (Braak >2 & ≤4, 10 subjects) and AD (Braak >4,8 subjects). Details of each case are described in Table 1. Hence, brains in the first group were considered as a (normal) control group (Braak ≤2). These were obtained from aged patients with no history of neurological disease and no pathological evidence of AD or other degenerative brain disease. The . Sections were first fixed by immersion in 4% paraformaldehyde (Sigma) buffer for 10 min, and then incubated in sodium borohydride buffer (1 mg/ml; Sigma) for 8 min to prevent auto-fluorescence. They were then incubated for 2 h at room temperature in blocking buffer containing 3% donkey serum and 0.1% triton X-100. Tissue sections were thereafter incubated overnight at 4°C with primary antibody (see Table 2) in 3% donkey serum and 0.1% triton X-100 containing buffer. The following day, the sections were washed and incubated 2 h at room temperature (in darkness) with Alexa Fluor â -conjugated secondary antibody diluted in 3% donkey serum and 0.1% triton X-100 at a concentration of 1:500. Slides were then cover-slipped using mounting medium (ProLong Gold antifade reagent, Life technologies; Thermo Fisher Scientific Inc.).
Avidin-biotin b-amyloid staining This procedure has been detailed in previous reports [37,38]. Briefly, sections of frontal and temporal cortex from the same patients used for ECM immunohistochemistry were cut at 6 lm thickness from formalin fixed, paraffin embedded blocks and mounted on to glass slides (SuperFrost Plus Gold â ; Thermo Fisher Scientific Inc.). Sections were firstly rehydrated in xylene followed by alcohols of decreasing concentration and distilled water. Sections were then subjected to chemical antigen retrieval (70% formic acid for 20 min at room temperature) and peroxidase activity was quenched (0.3% hydrogen peroxide in methanol for 30 min at room temperature). For all the following steps, horse serum (Vectastain Elite PK-6100; Vector Laboratories, Burlingame, CA, USA) was used as the blocking serum. The sections were next incubated for 1 h at room temperature in mouse monoclonal antibody directed against Amyloid-b 17-24 (4G8) (see Table 2) and then incubated with biotinylated secondary antibody for 30 min followed by 30 min incubation in ABC reagent (both Vectastain Elite PK-6100 mouse IgG). Sites of immunoreaction were visualized by

Microscopic analysis and image quantification
Images were acquired using a 109 objective of an Olympus BX51 microscope (Olympus, Tokyo, Japan) and a QImaging Retiga 6000 color camera system (Surrey, BC, Canada). For each staining, image quantification was performed on one section per brain area per patient; all staining being performed on adjacent sections. For each section of frontal and temporal cortices, snapshots of five random fields of view of cortical grey matter were taken for each immunostaining. Although fields selected for image quantification were chosen at random, these were always taken from within the same gyral region (superior frontal gyrus or inferior temporal gyrus) in each case. This sampling procedure provided a histological reference point ensuring that broadly the same region of cortex was quantified in each patient for each antibody used. As all the ECM molecules reported here are expressed at the level of the basal membrane or by endothelial cells and not in the parenchyma, we assume that all reactions pattern are most certainly representative of changes at vascular level. Image analysis was performed using ImageJ software (ImageJ, U.S. National Institutes of Health, Bethesda, MD, USA, http://imagej.nih.gov/ij/). Firstly, all snapshots were converted into 8-bit black and white images. Then, a threshold by intensity and by size of object was applied to each image to remove any non-specific immunostaining. Finally, results were summarized by averaging the total staining of the five snapshots for each antibody and expressed as area of positive staining per case and per mm 2 of tissue. Image acquisition and analysis were made blindly by a single observer. Higher magnification images (409) were taken to present the perivascular staining pattern of each stain and are shown in Figure S1.

Statistical analysis
A non-parametric one-way ANOVA (Kruskal-Wallis test) was used to compare the quantitative values of vascular staining for each antibody across the three groups in both frontal and temporal cortex. This was followed by post hoc analysis Dunn's test for multiple comparisons (the group sizes were not equal and all possible pairs were to be compared), and also to compare the mean age at death between groups. For correlations, the P-values in linear regression analysis were presented as well as the Pearson product-moment correlation coefficient r. Statistical significance was accepted at P < 0.05 level. Statistical tests were performed using GraphPad Prism version 6.00 for Windows (GraphPad Software, CA, USA, www.graphpad. com).

AD leads to changes in expression of ECM molecules
Subclinical AD and AD groups showed significantly higher levels of collagen IV ( Figure 1) and perlecan ( Figure 2) staining than the control group, for both frontal and temporal cortex, but no significant difference was observed between subclinical AD and AD groups. Likewise, similar changes to those observed for collagen IV and perlecan were seen for fibronectin ( Figure 3) (P < 0.05), although variability within groups was higher than for collagen IV and perlecan which led to no significant difference between the AD group and the control group in the temporal cortex.
Staining for hPECAM was weaker than the other ECM immunostaining. Nonetheless, quantification of hPECAM staining in frontal and temporal cortex did not show any significant differences between groups ( Figure 4). AD leads to equivalent Ab deposition between mild to moderate AD and moderate to severe AD groups Similar to collagen IV, perlecan and fibronectin, Ab deposition (represented by 4G8 immunostaining) was significantly higher in the subclinical AD and the AD groups vs. the control group in both frontal and temporal cortex (P < 0.001, for subclinical AD for both frontal and temporal cortex and for the AD group for frontal cortex, and P < 0.01, for the AD group for temporal cortex), but did not differ significantly between subclinical AD and AD groups in either area ( Figure 5).

Correlation between Ab deposition and ECM components
There was a significant correlation between 4G8 staining and collagen IV staining in both frontal and temporal cortex (Figure 6a) (P < 0.0001, r = 0.733 and P < 0.0001, r = 0.662, for frontal and temporal cortex, respectively), as well as between 4G8 and perlecan ( Figure 6b) (P = 0.0001, r = 0.641 and P < 0.0001, r = 0.712, for frontal and temporal cortex, respectively). Fibronectin and 4G8 staining were also correlated, but only in the temporal cortex ( Figure 6c) (P = 0.1516, r = 0.268 and P = 0.0223, r = 0.416, for frontal and temporal cortex, respectively). Finally, consistent with the lack of changes in hPECAM staining, no correlation between Ab deposition and hPECAM was seen ( Figure 6d) (P = 0.4552, r = 0.142 and P = 0.5912, r = 0.102, for frontal and temporal cortex, respectively).

AD leads to a decrease in vessel density
A significant decrease in vWF staining was detected in the temporal cortex (P < 0.05) (Figure 7) in the AD group when compared to the control and subclinical AD groups. A similar trend to decrease in vWF staining, although not significant (P = 0.0813), was observed in the frontal cortex when comparing the control group with the subclinical AD and AD groups.

Discussion
In the present study, we have found early (i.e. in individuals with mild to moderate AD-type pathology; subclinical AD group) as well as late (i.e. in patients with established AD; Braak >4 group) increases in expression levels of collagen IV, perlecan and fibronectin as observed by immunohistochemistry. It must be noted that all changes described in this work were not visible on the standard hematoxylin and eosin staining (data not shown) and were only detectable by immunohistochemistry. Furthermore, Ab accumulation is significantly increased in subclinical AD patients (when compared to control subjects) with no further increase in later Braak stages. This latter finding is not totally unexpected since the subclinical AD cases are likely to be cases of probable AD, even though they may not have met yet the pathological criteria for definite AD. The similar patterns of Ab accumulation and ECM changes logically lead to significant correlations between these parameters. Interestingly, we also show that these increase in ECM were not associated with an increase in vascular density, since vWF staining was actually unchanged in the frontal cortex and significantly decreased in temporal cortex of AD patients when compared to control subjects and subclinical AD.
The ECM provides a biochemical scaffolding that is essential for tissue elasticity, growth, remodelling, maintenance and stabilization of tissue structures and represents a critical interface between tissue compartments providing essential attachment, survival, migration and functional clues for cells [20]. Over the last decades, several studies have disclosed changes in cell surface proteoglycans and/or ECM components in AD. Proteoglycans such as agrin, glypicans and syndecans, have been reported to be associated with senile plaques and NFT [41][42][43]. Berzin and co-workers [44] have shown that soluble agrin levels were significantly increased from . Staining for a basement membrane-derived heparan sulphate proteoglycan was shown to be associated with diffuse plaques in both AD and Down's syndrome by Snow et al. [31], suggesting an early accumulation during plaque development. In an earlier study, Snow et al. [45], also demonstrated that proteoglycans colocalized with blood vessels in the brain parenchyma in non-affected areas while proteoglycan staining was lost in affected areas, suggesting a remodelling of the basement membrane and relocalization of its constituent during amyloidogenesis. Here, we show that AD is associated with modifications in ECM components, as shown by a higher level of staining for these proteins. Our findings are in agreement with a previous study, which found a 55% increase of collagen IV content in cerebral microvessels in AD patients, compared to controls [26]. Other studies in AD [46,47], and in transgenic animal models of AD [48,49], also reported a thickening of the basement membrane together with an increase in collagen content. Moreover, alterations in microvascular morphology have been described in AD with a decrease in the diameter of capillaries correlating with a deteriorating cognitive status [50]. Decreased cerebrovascular volume, higher levels of collagen IV and increased basement membrane thickness were also reported by Bourasset et al. [51] in the brains of 11-month-old triple (APP 9 PS1 9 Tau) transgenic mice. Additionally, the present data show higher levels in perlecan (as suggested by Kalaria et al. [26]) and fibronectin, which corroborates results from Lema nska-Perek and et al. [33] in mild to moderate AD pathology, and reinforces the idea of a thicker basement membrane. These data further support the argument that the possible reduction of the cerebrovascular volume observed by Bouras et al. [50] and Bourasset et al. [51] could be the result of a thickening of the brain microvessels, or more specifically of their basement membrane.
Various processes may cause such changes in ECM in AD. One possible explanation for these ECM accumulations might lie with a reduction in brain proteolytic systems. The natural tissue endogenous inhibitors (TIMPs), as well as the matrix metalloproteinases (MMPs), are implicated in the regulation of ECM metabolism, and changes in activity of these proteases may contribute to vascular remodelling by modulating ECM components [52,53]. Moreover, reports have shown an increase in levels of MMP-1 [54] or a decrease in MMP-9 [55] in brain, and Mroczko and co-workers [56] have shown an increase in MMP-3 and a decrease in MMP-9 in cerebrospinal fluid of AD patients. Dysregulation of these proteases might result in a decreased turnover of the basement membrane with a consecutive build-up of ECM components and potential dysfunction of the blood-brain barrier (BBB). In fact, an increasing amount of evidence suggests that MMPs and TIMPs may play a crucial role in the pathogenesis of AD [21,57,58]. These reports together with our findings warrant further studies to elucidate the role of these proteases in pathophysiology of AD. On the other hand, because grey matter atrophy has been previously reported in AD [59][60][61][62], we cannot rule out that some degree of grey matter shrinkage, leading to an increase in vessel density, could partly explain some of the increase in vessel components we have observed. However, this seems unlikely for several reasons. Firstly, our vWF staining data, a commonly used endothelial cell marker, suggests that late stage AD patients have no change, or even a decrease in vascular density in frontal and temporal cortex, respectively. Such decreases in capillary density are in agreement with previous studies [63][64][65][66]. Secondly, the magnitude of the component increase measured in the present study [+51% to 59% for collagen IV (see Figure 1), +23% to 79% for perlecan (see Figure 2) and +96% to 112% for fibronectin (see Figure 3)] is far greater than the degree of grey matter shrinkage, commonly reported to be around 5-15% in terms of volume in the frontal and temporal areas [59][60][61][62]. Finally, angiogenesis could be another explanation for an increase in vascular components, but to date there is only little evidence supporting this. For example, Desai and collaborators [67] showed an increase in integrin a V b 3 in the brain in AD, but this was associated with an increase in blood vessel density only in the hippocampus, an area which was not included in the present study. Accordingly, our first hypothesis based on a potential dysfunction of the proteolytic systems seems to be the most likely explanation for the increase of ECM components expression per vessel.
In our study, we did not find any difference in hPE-CAM staining between the different Braak stages. The hPECAM is a transmembrane immunoglobulin expressed by endothelial cells, where it is localized at the intercellular points of contacts, and by leucocytes and platelets [68], which in turn are known to play a role in Ab deposition and inflammatory processes [69]. Plasma levels of hPECAM were found to be higher in AD patients vs. controls [69][70][71], but no significant differences between the moderate to severe AD group and the mild AD group could be identified [71]. The difference between these reports and our present results may be due to a bias in our control group in which cases with Braak stage ≤2 and those with Braak 0 (i.e. 'true' control cases) were pooled. However, it is more likely that hPECAM is released into the circulation rather than accumulating in the basement membrane, hence explaining why plasma levels are higher whereas vascular levels remain fairly constant. Again, roles of the MMPs and TIMPs in the regulation of the expression of the junctional adhesion protein need to be considered [72].
Finally, we found positive, although modest (r = 0.41-0.773), correlations between Ab deposition and changes in ECM components. Overall, this result is in agreement with previous findings reporting an association between vascular Ab deposition and alterations of the cerebral vasculature such as smooth muscle loss [73] together with collagen IV deposition [30], apoptosis of cerebral vascular cells [74]. Altogether, these observations and previous reports support the concept that vascular changes are concomitant to disease progression, whether they are causative/contributing or simply consequences of the disease remains to be investigated.
In conclusion, our present results suggest that changes in ECM components could be potential biomarkers for investigational studies, at plasma level or using imaging techniques, in AD from early Braak stages, due to their alterations from Braak >2, or in animal models to determine the time-course of these changes at the vascular level and the potential causality between them and AD pathology. Indeed, the recur-rent problem is that the diagnosis of AD is based on a combination of cognitive tests, imaging and exclusion of other neurological disorders. Unfortunately, in most cases, the diagnosis is ascertained only when the disease has already been progressing for many years, and cases will likely exceed Braak stage 2 when diagnosed. Accordingly, our study supports the need to find molecular markers that will help diagnose AD as early as possible, as blood plasma measures [75], and to provide objective and reliable measures of disease progression. Imaging agents to measure ECM molecules in human patients could be useful, but these would need to access the basement membrane, which is beyond the BBB formed by tight junctions between endothelial cells, and therefore such approach would be very challenging. Plasma measures of ECM components seem better adapted for potential use as biomarkers to predict development of AD. Finally, it has been shown that ECM molecules may also represent potential thera-  The P-value as well as the Pearson product-moment correlation coefficient is shown in each graph. peutic targets for stroke or AD [76,77]; therefore our present data could be useful in defining new therapeutic strategies.