Involvement of matrix metalloproteinase-9 in amyloid-β 1-42-induced shedding of the pericyte proteoglycan NG2.

Deposition of amyloid-β (Aβ) 1-42, the major component of senile plaques characteristic of Alzheimer disease, affects brain microvascular integrity and causes blood-brain barrier dysfunction, increased angiogenesis, and pericyte degeneration. To understand the cellular events underlying Aβ1-42 effects on microvascular alterations, we investigated whether different aggregation forms of Aβ1-42 affect shedding of the pericyte proteoglycan NG2 and whether they affect proteolytic cleavage mediated by matrix metalloproteinase (MMP)-9. We found decreased levels of soluble NG2, total MMP-9, and MMP-9 activity in pericyte culture supernatants in response to fibril-enriched preparations of Aβ1-42. Conversely, oligomer-enriched preparations of Aβ1-42 increased soluble NG2 levels in the supernatants. This increase was ablated by the MMP-9/MMP-2 inhibitor SB-3CT. There was also a trend toward increased MMP-9 activity observed after oligomeric Aβ1-42 exposure. Our results, demonstrating an Aβ1-42 aggregation-dependent effect on levels of NG2 and MMP-9, support previous studies showing an impact of Aβ1-42 on vascular integrity and thereby add to our understanding of mechanisms behind the microvascular changes commonly found in patients with Alzheimer disease.

INTRODUCTION

Pericytes play a vital role in brain blood vessel function and integrity. Wrapped around the wall of capillaries, they stabilize the blood vessels and regulate blood flow by responding to vasoactive substances (1). Pericytes are also, as one of the key players in the neurovascular unit, important for the maintenance of the blood brain barrier (BBB) (2, 3), which protects the brain from periphery-derived harmful substances and controls influx of nutrients. This cell-type further expresses several BBB-related transporters, which implicates these cells in brain-to-blood efflux (4, 5). Since pericytes also have phagocytic properties (6), they are regarded as important actors in the clearance of neurotoxic substances from the brain parenchyma. In addition, pericytes regulate the formation of new blood vessels as they secrete angiogenic factors, for example, vascular endothelial growth factor (7) and matrix metalloproteinases (MMPs) (8, 9); MMPs proteolytically degrade the extracellular matrix, paving the way for migrating cells and sprouting blood vessels (10).

Recent studies have implicated pericytes in the pathogenesis of Alzheimer disease (AD) (11–13), which is foremost associated with a progressive accumulation of aggregated amyloid-β (Aβ) 1–42 and intraneuronal neurofibrillary tangles of hyperphosphorylated tau (14). The brains of AD patients also exhibit neuroinflammation (15), white matter changes (16, 17), and microvascular pathology. Microvascular abnormalities include BBB dysfunction, aberrant angiogenesis, altered microvascular density, and atrophic and ruptured microvessels (12, 13, 18–21)—events that may involve pericyte malfunction. Indeed, a recent study demonstrated an approximately 60% reduction of pericytes and blood vessel coverage in AD patients versus neurologically intact controls. The reduction in pericytes was shown to correlate with BBB dysfunction and amyloid plaque load (22). The involvement of pericytes in AD-associated pathology has been demonstrated as pericyte loss in transgenic mice overexpressing the Aβ precursor protein (APP); the mice also have elevated, impaired and accelerated Aβ clearance (11). Moreover, cultured human pericytes have decreased viability after prolonged exposure to the Aβ peptide (23, 24). Thus, changes in the pericyte population, possibly caused by the increasing load of toxic Aβ, may underlie some of the microvascular events and the impaired clearance of Aβ with accelerating neurodegenerative processes as downstream events (1).

Pericytes are recognized by their expression of the transmembrane proteoglycan NG2 (25, 26). This proteoglycan plays an important role in the interaction between endothelial cells and pericytes. Specific targeting of NG2 using anti-NG2 antibodies inhibits ischemia-induced retinal angiogenesis in mice (27). Decreased pericyte ensheathment of endothelial cells, diminished formation of endothelial cells, and reduced assembly of the vascular basal lamina have been demonstrated in pericyte-specific NG2-null mice (28). In vitro studies have also demonstrated that NG2 knockdown in pericytes inhibits 3-dimensional pericyte–endothelial cell network formation, pericyte proliferation/migration, and endothelial cell junction formation, which leads to increased endothelial cell monolayer permeability (28). Interestingly, shed soluble NG2 (sNG2) has been shown to promote angiogenesis and migration of endothelial cells via binding of sNG2 to ligands expressed on endothelial cells (29).

We have previously demonstrated decreased levels of sNG2 and a correlation between sNG2 and Aβ1–42 in the cerebrospinal fluid (CSF) of patients with AD (30). We have also shown decreased CSF sNG2 levels in patients with dementia with Lewy bodies, but not in patients with Parkinson disease (31). Both of these disorders are neuropathologically characterized by abnormal intraneuronal inclusions of alpha (α)-synuclein (Lewy bodies), but DLB patients may also display Aβ plaques similar to AD patients (32). We further demonstrated a direct inhibiting effect of fibrillar Aβ1-42 on NG2 shedding from cultured primary human oligodendrocyte precursor cells (OPC) (30), which also express NG2 (33). The mechanisms behind Aβ-induced NG2 shedding are not known, but MMPs, in particular MMP-9, have been shown to affect NG2 shedding in general. Broad-spectrum MMP inhibitors have been reported to decrease the amount of sNG2 in conditioned media from cultured rat oligodendrocyte precursor cells (34), and decreased sNG2 levels in rat spinal cord after infusion of the MMP-9/MMP-2 inhibitor SB-3CT have also been demonstrated (35). It has further been shown that NG2 is a substrate for MMP-9 (36) and that clearance of NG2 from injured brain areas is lost in MMP-9–null mice (36).

Whether Aβ1–42 also affects pericyte NG2 shedding and whether MMP-9 is the sheddase involved are not known. Here, we analyzed sNG2 levels in primary human pericyte culture supernatants after exposure to fibril- and oligomer-enriched Aβ1–42 preparations in the presence or in the absence of the MMP-9/MMP-2 inhibitor SB-3CT. We next measured the activities of MMP-9 and MMP-2 after exposure to fibrillar and oligomeric Aβ1–42. Lastly, we analyzed the impact of Aβ1–42 on cell viability and total concentrations of MMP-9, MMP-1, and MMP-3; the two latter MPPs are known to affect MMP-9 production and activity (37–39).

MATERIALS AND METHODS

Cells

Primary human brain vascular pericytes (HBVPs) isolated from fetal brain tissue (ScienCell Research Laboratories, San Diego, CA) were cultured in pericyte culture media (ScienCell Research Laboratories) containing 2% fetal bovine serum. Cells were grown as monolayers in poly-L-lysine–coated culture flasks in humidified air with 5% CO2 at 37°C until they were 80% to 90% confluent. Before the experiments, the cells were plated onto either 12-well chamber slides (Nunc A/S) or 8-well chamber slides (Lab-Tek) and grown until 80% confluent. Cell population purity was assessed by immunofluorescence staining using antibodies against NG2 (1:200; Millipore, Billerica, MA), platelet-derived growth factor receptor-β (PDGFR-β) (1:200; Abcam, Cambridge, MA), and α-smooth muscle actin (αSMA) (Sigma, St Louis, MO). Cells were fixed in 2% formaldehyde, incubated with blocking solution (PBS containing 1% BSA [Boehringer Mannheim, Vienna, Austria] and 5% goat serum [Jackson ImmunoResearch, West Grove, PA]) and thereafter incubated with the primary antibodies (anti-NG2 and anti-PDGFR-β in blocking solution) followed by an incubation with the appropriate secondary antibodies Alexa 488-conjugated anti-rabbit immunoglobulin G (IgG) (1:500 Molecular Probes) or Cy3-conjugated antimouse IgG (1:500, Jackson ImmunoResearch). Cells stained against SMA were fixed in 2% formaldehyde, permeabilized with 0.1% Saponin (Sigma-Aldrich) in 1% bovine serum albumin (BSA)/PBS, incubated with blocking solution, and incubated with Cy3-conjugated antibodies directed against anti-αSMA (1:200, Sigma). Finally, cells were mounted with Vectashield Set mounting medium with DAPI (Vector Laboratories, Burlingame, CA). The morphology of the cells corresponded well to previously published images of HBVP (40, 41); the percentage of DAPI-positive/NG2-positive/PDGFR-β–positive cells of total DAPI-positive cells (n = 250) was approximately 99% (Fig. 1). In line with previous studies on primary pericytes isolated from human and rodent brain tissues (42, 43), few (<1%) of the HBVPs expressed αSMA, indicating an undifferentiated state of the HBVPs.

FIGURE 1

Fetal HBVPs express the pericyte protein NG2 and PDGFR-β. Human brain vascular pericyte stained with DAPI (A), anti-NG2 antibody (B), and anti–PDGFR-β antibody (C). Images are merged in (D). Scale bar = 10 μm.

Aβ Preparations

Aβ1–42 oligomer- and fibril-enriched preparations were prepared by dissolving Aβ1–42 peptide (Bachem, Bubendorf, Switzerland) in cold hexaflouro-2-propanol (Sigma-Aldrich/Fluka), aliquoted, dried by speed vacuum, and stored at −80°C until used. Before cell treatment, hexaflouro-2-propanol/Aβ1–42 was dissolved in dimethyl sulfoxide (DMSO) to a final concentration of 2.5 mmol/L and sonicated for 10 minutes. For oligomeric preparation, Aβ1–42/DMSO was diluted to 100 μmol/L in phenol red–free Dulbecco modified Eagle medium and incubated for 24 hours at 4°C. Fibrillar preparations were generated by diluting Aβ1–42/DMSO to 100 μmol/L in 10 mmol/L HCl then incubated for 24 hours at 37°C. Differences between fibril-enriched and oligomer-enriched Aβ1–42 preparations before and after treatment of HBVPs were documented by Western blot analysis (Figure, Supplemental Digital Content 1, http://links.lww.com/NEN/A602). The Aβ1–42 preparations were separated on 10% Tris-tricine gels and transferred to polyvinylidene difluoride membranes (Millipore) using a transblot semidry transfer cell (BioRad, Hercules, CA) at room temperature for 60 minutes at a current of 50 mA. The membranes were rinsed, incubated with blocking solution (PBS-Tween containing 5% milk powder; Fluka) for 1 hour, and incubated with mouse anti–Aβ1–16 6E10 antibody (1:5000; Covance, Princeton, NJ) in blocking solution on a shaker at 4°C. The membranes were rinsed and incubated with anti-mouse Na931V antibody (1:1000; GE Healthcare, Deigen, Belgium) for 1 hour at room temperature on a shaker. Proteins were visualized using the Luminata Forte Western HRP Substrate (Millipore) and a LAS-3000 charge-coupled device camera (FujiFilm).

HBVP Treatment

Before cell treatment, the pericyte culture media were removed and replaced with serum-free culture media (ScienCell Research Laboratories) containing 10 μmol/L Aβ1–42 fibril-enriched preparations and 10 μmol/L Aβ1–42 oligomer-enriched preparations. Cells were treated with DMSO/phenol red–free Dulbecco modified Eagle medium (vehicle for Aβ1–42 oligomer-enriched preparations) and DMSO/HCl (vehicle for Aβ1–42 fibril-enriched preparations) as controls. Potential involvement of MMP-9 in Aβ1–42–regulated NG2 shedding was analyzed by adding 10 μmol/L of the MMP-9/MMP-2 inhibitor SB-3CT (Calbiochem, Billerica, MA) to HBVPs exposed to vehicles, 10 μmol/L Aβ1–42 fibril-enriched preparations, or 10 μmol/L Aβ1–42 oligomer-enriched preparations. The cells were incubated for 24 hours at 37°C. All experiments were performed in duplicate and repeated independently 3 times. Cell culture supernatants were collected after treatment, centrifuged (275 × g, 5 minutes, 4°C), aliquoted, and stored at −80°C until used. The cells were rinsed with PBS, lysed with the mammalian Cell Lysis Kit (Sigma-Aldrich) according to the manufacturer’s protocol, aliquoted, and stored at −80°C until used. Similar protein concentrations in the wells with different treatment conditions were verified with the Bradford (Coomassie plus) assay kit (Thermo Scientific).

NG2 Quantification Assay

Levels of NG2 in HBVP cell culture supernatants and lysates were analyzed on an in-house–developed assay with Meso Scale Discovery electrochemiluminescence technology using immunoassay conversion kits (Meso Scale Discovery, Rockville, MD). Cell culture supernatants and lysates were diluted (1:5 and 1:250, respectively) in PBS and coated onto Meso Scale Discovery high-bind plates in triplicate wells (25 μL) and allowed to adhere overnight at 4°C with slow agitation. The wells were thereafter washed with washing buffer (0.05% Tween 20 in PBS) and incubated with blocking solution (1% bovine serum albumin and 1% milk in PBS-Tween) for 1 hour. After rinse, the wells were incubated with mouse anti-NG2 (1:100, ATCC clone B5; kind gift from Dr William Stallcup, Burnham Institute, La Jolla, CA) for 2 hours on a shaker, washed, and incubated with goat anti-mouse sulftag-conjugated secondary antibodies (1:500; Meso Scale Discovery) for 1 hour on a shaker. The electrochemiluminescence signal was quantified using a Meso Scale Discovery SECTOR Imager 6000.

MMP-9 and MMP-2 Activity Assays

SensoLyte Plus 520 MMP-9 and MMP-2 Assay Kits (AnaSpec, Fremont, CA) were used according to the manufacturer’s instructions to quantify MMP-9 and MMP-2 activities in HBVP cell culture supernatants. The assays are MMP-9– and MMP-2–specific, with no cross-reactivity with other MMPs. The MMP activity unit was expressed as the change in fluorescence intensity (relative fluorescence unit) at an excitation of 490 nm and at an emission of 520 nm. The emission at 520 nm was measured after 1 hour using the Microplate Spectophotometer Infinite M200 and the Magellan version 3.5 software.

MMP Quantification

Human MMP 3-Plex Ultra-Sensitive immunoassay (Meso Scale Discovery, Oxford, United Kingdom) was used according to the manufacturer’s protocol to quantify both the active and the proenzyme forms of MMP-1, MMP-3, and MMP-9. Detection limits for MMP-1, MMP-3, and MMP-9 in HBVP cell culture supernatants were 3.15, 3.39, and 11.8 pg/mL, respectively.

Lactic Dehydrogenase Cytotoxicity Assay

Cell viability was determined as a measure of lactic dehydrogenase (LDH) activity in HBVP cell culture supernatants, as previously described (30, 44). Supernatants from control cell cultures treated with vehicle for Aβ1–42 fibril-enriched or oligomer-enriched preparations were used as baseline controls; supernatants from cells treated with 1 μmol/L staurosporine (Sigma-Aldrich) for 24 hours were used as positive controls for cytotoxicity. K values obtained were averaged (mean ± SD) and expressed as percentage of activity in cell culture supernatants from baseline controls.

Statistical Analysis

Statistical analysis was performed using the SPSS software (version 20 for Windows; SPSS Inc, Chicago, IL). To detect significant differences between the different experimental conditions, we analyzed the data using paired t-test. Results are presented as mean ± SD. A value of p < 0.05 was considered significant.

RESULTS

Aβ1–42 Alters sNG2 Levels in an Aggregation-Dependent Manner

Initially, we investigated whether fibril- and oligomer-enriched preparations of Aβ1–42 differently affect HBVP levels of surface-bound or shed NG2. Analysis of HBVP lysates demonstrated unaltered NG2 levels after exposure to both oligomeric and fibrillar Aβ1–42 versus vehicle (Fig. 2A; Table). Levels of sNG2 in HBVP culture supernatants were, however, significantly decreased compared with vehicle control upon exposure to fibrillar Aβ1–42 (Fig. 2B; Table). In contrast, sNG2 levels were significantly increased compared with vehicle upon oligomeric Aβ1–42 exposure (p = 0.036) (Fig. 2B; Table).

FIGURE 2

Bar graphs demonstrating alterations in the levels of the pericyte protein NG2 in cell culture supernatants and lysates from HBVPs after exposure to 10 μmol/L oligomer- or fibril-enriched Aβ 1–42 preparations for 24 hours in the presence or in the absence of the MMP-9/MMP-2 inhibitor SB-3CT. (A) No alterations in NG2 levels were found in HBVP lysates after exposure to either fibril- or oligomer-enriched Aβ1–42 preparations. (B) Exposure to fibrillar Aβ1–42 significantly decreased, whereas oligomeric Aβ1–42 significantly increased, the levels of sNG2 (% of vehicle). (C) The increased levels of sNG2 in HBVP cell culture supernatants observed after exposure to oligomer-enriched Aβ1–42 preparations were inhibited in the presence of SB-3CT. Data were analyzed using paired t-test. * Significant difference at p < 0.05 level versus baseline. ** Significant difference at p < 0.01 level versus baseline.

TABLE

Variables Analyzed in HBVP Lysate and Cell Culture Supernatants After Aβ1–42 Exposure

SB-3CT Ablates Oligomeric Aβ1–42–Induced NG2 Shedding

Previous studies have shown decreased levels of sNG2 in rat spinal cord after infusion of the MMP-9/MMP-2 inhibitor SB-3CT. Therefore, we investigated whether SB-3CT affects Aβ-induced NG2 shedding. In the presence of SB-3CT, the levels of sNG2 from HBVPs exposed to fibrillar Aβ1–42 remained significantly decreased versus those for vehicle (Fig. 2C; Table). The described increase in sNG2 levels after exposure to oligomeric Aβ1–42 was ablated in the presence of SB-3CT (Fig. 2C; Table).

Aβ1–42 Alters the Activity of MMP-9, But Not MMP-2, in an Aggregation-Dependent Manner

We next analyzed the activities of MMP-9 and MMP-2 when inhibited by SB-3CT in HBVP culture supernatants after exposure to oligomeric and fibrillar Aβ1–42. There was a significant decrease in MMP-9 activity after exposure to fibrillar Aβ1–42 versus vehicle (Fig. 3A; Table). On the contrary, oligomeric Aβ1–42 increased MMP-9 activity; however, although the results from all 3 independent experiments consistently showed higher MMP-9 activity in oligomeric Aβ1–42–exposed HBVPs compared with the respective vehicle, this increase did not reach significance (Fig. 3A; Table). Activity of MMP-2 did not change after exposure to either oligomeric or fibrillar Aβ1–42 compared with vehicle (Fig. 3B; Table).

FIGURE 3

Bar graphs demonstrating alterations in total MMP-9 and MMP-2 activities in cell culture supernatants from HBVPs after exposure to 10 μmol/L oligomer- or fibril-enriched Aβ 1–42 preparations for 24 hours. (A) Exposure to fibrillar Aβ1–42 significantly decreased MMP-9 activity (% of vehicle), whereas no significant change was found after oligomeric Aβ1–42 exposure. (B) No alterations in MMP-2 activity were found in HBVP cell culture supernatants after exposure to either fibrillar or oligomeric Aβ1–42. Data were analyzed with paired t-test. * Significant difference at p < 0.05 level versus baseline.

Aβ1–42 Alters MMP-9 and MMP-1 Levels in an Aggregation-Dependent Manner

Because our results support the notion of MMP-9 as a candidate sheddase involved in Aβ1–42–altered NG2 shedding, we measured total protein levels (active enzyme and proenzyme) of MMP-9 in Aβ1–42–exposed HBVP culture supernatants. Although MMP-9 concentrations remained unaltered upon treatment with oligomeric Aβ1–42, there were significantly decreased MMP-9 concentrations after fibrillar Aβ1–42 exposure (Fig. 4A; Table). We also measured total levels of MMP-3 and MMP-1—MMPs known to affect MMP-9 activity and production (37–39). Total MMP-1 concentrations were, like those of MMP-9, significantly reduced in the presence of fibrillar Aβ1–42, but not oligomeric Aβ1–42, compared with vehicle (Fig. 4B; Table). By contrast, significantly reduced levels of MMP-3 were found after oligomeric Aβ1–42, but not fibrillar Aβ1–42, exposure (Fig. 4C; Table).

FIGURE 4

Bar graphs demonstrating alterations in total MMP-9, MMP-1, and MMP-3 levels in cell culture supernatants from HBVPs after exposure to 10 μmol/L oligomer- or fibril-enriched Aβ 1–42 preparations for 24 hours. (A) Exposure to fibrillar Aβ1–42 significantly decreased total MMP-9 levels (% of vehicle), whereas no change in MMP-9 levels was found after oligomeric Aβ1–42 exposure. (B) Levels of total MMP-1 (% of vehicle) were significantly reduced after fibrillar Aβ1–42 exposure but were unaltered after oligomeric Aβ1–42 exposure. (C) There was a significant reduction in total MMP-3 levels (% of vehicle) after oligomeric Aβ1–42 exposure, but significance was not reached after fibrillar Aβ1–42 exposure. Data were analyzed with paired t-test. * Significant difference at p < 0.05 level versus baseline. ** Significant difference at p < 0.01 level versus baseline.

Aβ1–42 Exposure Does Not Alter Pericyte Viability

To rule out the possibility that the observed alterations in sNG2 levels resulted from Aβ1–42– induced cell death, we measured extracellular LDH activity (as an indication of cell death) in culture supernatants from HBVPs exposed to oligomeric and fibrillar Aβ1–42. As expected, cells exposed to the positive control staurosporine demonstrated significantly increased extracellular LDH activity versus vehicle-treated cells, but no alterations in LDH activity were found after exposure to either fibrillar or oligomeric Aβ1–42 (Fig. 5).

FIGURE 5

Bar graphs demonstrating alterations in LDH activity, indicative of cell viability, in cell culture supernatants from HBVPs after exposure to 10 μmol/L oligomer- or fibril-enriched Aβ1–42 preparations for staurosporine for 24 hours. Viability of HBVP was not altered after exposure to oligomeric or fibrillar Aβ1–42 but was significantly increased in response to the positive control staurosporine. Data were analyzed with paired t-test. * Significant difference at p < 0.05 level versus baseline.

DISCUSSION

We recently demonstrated decreased levels of sNG2 in the CSF of patients with AD (30) and dementia with Lewy bodies (31), both of which are associated with extracellular deposition of Aβ. Our previous results also demonstrated decreased NG2 immunoreactivity in the vicinity of amyloid plaques in individuals with significant plaque loads. We further showed that the levels of sNG2 are decreased in culture supernatants from NG2-expressing HOPCs exposed to fibrillar Aβ1–42 (30). Extending our previous findings, we show here that the Aβ-induced decrease in NG2 is not cell-specific because fibrillar Aβ1–42 also significantly decreased sNG2 levels in the culture supernatant from primary human pericytes. We also show that in contrast to fibrillar Aβ1-42, oligomeric Aβ1‐42 significantly increased sNG2 levels in cultured pericytes, whereas the contents of the HBVP lysate NG2 remained unaltered. Together, these results demonstrate an aggregation-dependent effect of Aβ1–42 on NG2 shedding from the cell surface.

Interestingly, the oligomeric Aβ1–42–induced increase in sNG2 was ablated in the presence of SB-3CT. Activity of MMP-9 was significantly lower after fibrillar Aβ1–42 exposure, whereas a slight increase in activity after oligomeric Aβ1–42 exposure did not reach significance. No alteration in MMP-2 activity was detected after exposure to either Aβ1–42 aggregation form. Fibrillar Aβ1–42 furthermore decreased the total levels of MMP-9 and MMP-1, but not MMP-3, whereas oligomeric Aβ1–42 did not affect the levels of MMP-9 or MMP-1 but significantly decreased the level of MMP-3. Finally, neither Aβ1–42 aggregation form affected HBVP viability.

In our previous study on HOPCs, we did not detect any difference in sNG2 levels after oligomeric Aβ1–42 exposure (30). Thus, as previously suggested by Asher et al (34), the presence of active NG2 sheddase is cell type–dependent (i.e. the enzyme responsible for Aβ1–42–induced enhanced proteolytic cleavage of NG2 in HBVP cultures may not be secreted or is inhibited in HOPC cultures). This hypothesis is strengthened by the fact that Asher et al (34) did not find an inhibitory effect on rat oligodendrocyte precursor cell NG2 shedding in the presence of metallopeptidase inhibitor-1 (TIMP-1), an inhibitor that acts on MMPs, including MMP-9. In the present study, we demonstrated an inhibitory effect of the MMP-9/MMP-2 inhibitor SB-3CT on oligomeric Aβ-induced NG2 shedding.

To determine whether the effect of SB-3CT was caused by an inhibitory effect on MMP-9 or MMP-2 activity, we measured the activities of both MMPs after Aβ1–42 exposure. In agreement with the altered sNG2 levels, we found decreased MMP-9 activity after fibrillar Aβ1–42 exposure and a trend toward increased MMP-9 activity after oligomeric Aβ1–42 exposure. By contrast, MMP-2 activity was unaltered after exposure to either aggregation variant. These results support the idea that an Aβ1–42–induced alteration in NG2 shedding is mediated specifically via the MMP-9 pathway and not via the MMP-2 pathway.

Our analysis of total MMP levels (i.e. proenzymes, active enzymes, and enzyme/inhibitor complexes) again showed a decrease in MMP-9 after fibrillar Aβ1–42 exposure, indicating a decrease in MMP-9 secretion in response to this Aβ1–42 aggregation form. Surprisingly, although oligomeric Aβ1–42 stimulation increased MMP-9 activity, it did not increase total MMP-9 levels. This incongruence suggests that secretion of MMP-9 is not altered in response to oligomeric Aβ1–42; rather, there is a shift in the proenzyme/active enzyme ratio or a shift in activated enzyme/inhibited enzyme that underlies the unaltered total MMP-9 level. This result further underscores the importance of carefully differentiating the interpretation of different MMP-9 assays.

Our total MMP analysis also demonstrated an aggregation-dependent impact of Aβ1–42 on total levels of MMP-1 and MMP-3. Both MMPs have been shown to enhance MMP-9 expression in macrophages (39), and MMP-3 is known to activate the proenzyme MMP-9 (37, 38, 45). Our study instead showed that secretion of MMP-3 in response to the 2 Aβ1–42 preparations was altered in the opposite direction compared with secretion of MMP-9 and MMP-1. Although this result is not in line with the previous study on macrophages, our study is in agreement with a clinical study demonstrating decreased levels of MMP-9 and increased levels of MMP-3 in plasma from patients with AD (46).

The aggregation-dependent impact of Aβ1–42 on NG2 shedding and MMP secretion is highly relevant in view of the increasing number of studies supporting the idea that soluble oligomeric—not fibrillar—Aβ1–42 is the causative agent in the pathogenesis of AD (47, 48). For example, studies on the Tg2576 AD mouse model have shown hypervascularization, tight junction disruption, and BBB dysfunction after an early rise in amyloid precursor protein expression even before the onset of cognitive deficits and the presence of senile plaques (49, 50). In addition, active Aβ immunization was shown to a reverse these microvascular pathologic events (51, 52). These findings propose an impact of Aβ on BBB integrity even at the early stages of AD pathology and, more importantly, suggest that the soluble oligomeric Aβ1–42 or other Aβ species (rather than the senile plaque-forming fibrillar Aβ1–42) may be the culprits in Aβ-mediated BBB dysfunction. Our study demonstrates a significant increase in NG2 shedding in response to high levels of oligomeric Aβ1–42, in line with the concept previously presented by Biron et al (49). Based on the notion that amyloidogenesis in part underlies the microvascular pathology observed at the early stages of AD, the previously reported involvement of NG2 in angiogenesis and BBB integrity, and our findings in the current study, it is tempting to hypothesize around the occurrence of the following scenario (Fig. 6): At the early stages of AD pathogenesis (i.e. before symptoms and cognitive decline become evident in patients with AD), pericytes respond to elevated levels of toxic oligomeric Aβ1–42 by secreting less MMP-9 inhibitors. Increased MMP-9 activity would lead to enhanced degradation of the extracellular matrix and tight junction proteins, thereby paving the way for cell migration across the BBB and along the vasculature (10, 53, 54). As a consequence, the MMP-9 substrate NG2 is shed from the cell surface of pericytes, thereby contributing to a loss of junctional interaction between pericytes and endothelial cells and between endothelial tight junctions (28, 55), leading to a dysfunctional BBB. The increased availability of sNG2 levels would also promote angiogenic activity (29) and, in combination with increased BBB permeability (56, 57) and degeneration of pericytes (23, 24, 58), blood vessels would be rendered unstable, disrupted, and leaky, which would eventually lead to microbleeds and incomplete infarcts (59, 60). As the pathology progresses, fibrillar Aβ1–42 accumulates and inhibits the cleavage of NG2 (30). This event could possibly be viewed as a protective mechanism, but the drastic loss (i.e. below normal levels) of available sNG2 may also reflect the loss of a factor involved in several crucial cellular events, including neuronal guidance, cell proliferation, and cell migration (26, 61). Support for the involvement of MMP-9 in the early pathogenesis of AD can be found in a previous clinical biomarker study demonstrating increased MMP-9 in the CSF of cognitively healthy elders with risk markers for future AD (62). Moreover, BBB dysfunction in mice carrying the APOEε4 genotype (a strong genetic risk factor for AD) was previously shown to be mediated by the proinflammatory cytokine cyclophilin A, which in turn promotes MMP-9 production. Remarkably, the BBB impairment in these mice could be reversed by SB-3CT treatment and specific inhibition of MMP-9, but not of MMP-2 (63). Whether CSF NG2 levels are altered at the early stages of AD pathogenesis remains to be investigated, but the previously reported decrease in CSF NG2 levels in AD patients (30) could possibly result from the inhibitory impact of fibrillar Aβ1–42 observed in our studies. Notably, the NG2 research field is still in its infancy, and we lack direct evidence for an impact of pericyte NG2 shedding on the vascular integrity of the BBB. We would, therefore, emphasize the importance of further investigations into the role of NG2 shedding in BBB integrity, preferably using experimental in vivo rodent models or clinical studies.

FIGURE 6

Image illustrating the hypothesized role of sNG2 in Aβ-induced microvascular pathology, as follows: 1) in response to increased levels of toxic oligomeric Aβ; 2) pericytes secrete extracellular MMPs, including MMP-9; 3) increased activated MMP-9 enhances NG2 shedding; 4) loss of extracellular matrix resulting from MMP-9 proteolysis causes pericytes to detach from the blood vessel; 5) the increased amount of sNG2 induces proliferation of endothelial cells; 6) BBB permeability increases; 7) extended exposure to Aβ degenerates pericytes; 8) lost pericyte support and fragility lead to rupture of microvessels.

In conclusion, the results from our study demonstrate an aggregation-dependent impact of Aβ1–42 on the shedding of pericyte NG2, a proteoglycan known to play a regulatory role in the maintenance of vascular integrity. Moreover, our study highlights MMP-9 as one of the mediators of Aβ1–42–induced alteration in NG2 shedding. These results are in line with the hypothesis suggesting amyloidogenesis as an event implicated in the microvascular pathology commonly found at the early stages of AD pathogenesis and highlight pericytes (possibly via the NG2 proteoglycan) as potential players in the events occurring in early AD pathogenesis.