Matrix Metalloproteinase-9 Degrades Amyloid-β Fibrils in Vitro and Compact Plaques in Situ*

The pathological hallmark of Alzheimer disease is the senile plaque principally composed of tightly aggregated amyloid-β fibrils (fAβ), which are thought to be resistant to degradation and clearance. In this study, we explored whether proteases capable of degrading soluble Aβ (sAβ) could degrade fAβ as well. We demonstrate that matrix metalloproteinase-9 (MMP-9) can degrade fAβ and that this ability is not shared by other sAβ-degrading enzymes examined, including endothelin-converting enzyme, insulin-degrading enzyme, and neprilysin. fAβ was decreased in samples incubated with MMP-9 compared with other proteases, assessed using thioflavin-T. Furthermore, fAβ breakdown with MMP-9 but not with other proteases was demonstrated by transmission electron microscopy. Proteolytic digests of purified fAβ were analyzed with matrix-assisted laser desorption ionization time-of-flight mass spectrometry to identify sites of Aβ that are cleaved during its degradation. Only MMP-9 digests contained fragments (Aβ1-20 and Aβ1-30) from fAβ1-42 substrate; the corresponding cleavage sites are thought to be important for β-pleated sheet formation. To determine whether MMP-9 can degrade plaques formed in vivo, fresh brain slices from aged APP/PS1 mice were incubated with proteases. MMP-9 digestion resulted in a decrease in thioflavin-S (ThS) staining. Consistent with a role for endogenous MMP-9 in this process in vivo, MMP-9 immunoreactivity was detected in astrocytes surrounding amyloid plaques in the brains of aged APP/PS1 and APPsw mice, and increased MMP activity was selectively observed in compact ThS-positive plaques. These findings suggest that MMP-9 can degrade fAβ and may contribute to ongoing clearance of plaques from amyloid-laden brains.

One of the key pathological features of Alzheimer disease (AD) 3 is the senile plaque, extracellular deposits found throughout the brains of AD patients, composed primarily of the amyloid-␤ peptide (A␤). Aggregated A␤ in senile plaques can be found in two conformations: non-␤-pleated-sheet (nonfibrillar) conformation, known as "diffuse plaques" or ␤-pleated sheet (fibrillar) conformation, known as "compact plaques" (1).
The 42-amino acid peptide (A␤ , the predominant peptide length found in senile plaques, has a remarkable propensity to aggregate at high concentrations to form a ␤-pleated sheet structure (2,3). This is generally viewed as an irreversible process resulting in the formation of fibrillar A␤ (fA␤), which is insoluble and resistant to proteolysis, endowing compact plaques with a resistance to degradation and clearance. Given the purported irreversibility of A␤ fibril formation and the resistance to degradation, one might expect that senile plaques would continue to grow throughout disease progression; however, careful observational studies indicate that plaque size remains relatively constant over a wide range of disease durations (4). In addition, in vivo imaging in the APPsw (Tg2576) transgenic mouse model of Alzheimer's disease (using multiphoton microscopy) demonstrated that plaques remain constant in size over a period of many months (5). These observations have led some to believe that plaques, once formed, are in dynamic equilibrium with their environment, balancing formation with degradation (6). In support of this idea, a few isolated plaques in APPsw mice were found to decrease in size (5), raising the possibility that endogenous mechanisms for plaque clearance may exist. Moreover, a semiquantitative analysis of plaque burden in Alzheimer disease cases revealed that the most advanced cases (by Braak staging) actually had a slightly lower frequency of plaques than less advanced cases (7).
While plaques and amyloid fibrils have been viewed by some as resistant to proteolytic degradation, it is possible that certain (as yet unidentified) proteases may contribute to endogenous mechanisms leading to plaque clearance. A growing list of proteases are known to degrade soluble A␤ (sA␤) in vitro, including NEP (8), IDE (9), ECE (10), angiotensin-converting enzyme (11), the plasmin system (12), and MMP-9 (13). Each of these proteases appears to have distinguishing A␤ cleavage sites, generating characteristic fragments. However, experimental evidence that these proteases degrade sA␤ in vivo is available only for a few: gene deletion of NEP (14), ECE (15), or IDE (16) in mice resulted in increased steady-state levels of A␤ in brain, suggesting a role in regulating endogenous basal levels of A␤ in vivo. Although all of these proteases demonstrate the ability to degrade sA␤, none has been convincingly shown to degrade A␤ fibrils or compact plaques.
MMP-9, a zinc-dependent metalloprotease, has not been well studied with regards to its potential role in amyloid clearance. Initially synthesized as an inactive proenzyme, pro-MMP-9 is cleaved into an active form upon cellular release by other proteases (17); this property puts MMP-9 in a unique position to regulate extracellular A␤ levels. Once activated, MMP-9 can be regulated by interacting with endogenous tissue inhibitors of MMPs (TIMPs) (17). Although expressed at low basal levels in normal brain tissue, MMP-9 expression is induced in neurons, astrocytes, microglia, and vascular cells under a variety of pathological conditions (18 -23). Its expression is increased in the brains of AD patients (13,24). Moreover MMP-9 expression in astrocytes is induced in the presence of the A␤ peptide (25).
In this study, we investigated the possibility that certain sA␤degrading proteases were capable of degrading fA␤ and amyloid plaques. Specifically, we examined the well studied sA␤degrading proteases, ECE, IDE, and NEP, as well as MMP-9. We further examined the expression and activity of fA␤-degrading proteases in the brains of transgenic mouse models of AD.

EXPERIMENTAL PROCEDURES
APP Transgenic Mice-Tg2576 (APPsw) and APP/PS1 mice, models of AD, were used in this study. The production, genotyping, and background strains of these mice have been described previously (26). Transgenic mice were compared with age-matched littermate wild-type controls. All experimental protocols were approved by the Animal Studies Committee at Washington University.
A␤ Preparation and Analysis-Synthetic human A␤ 1-42 (Bachem) or A␤ 1-40 (American Peptide), dissolved in dimethyl sulfoxide (Me 2 SO, Sigma) to a concentration of 5 mM, was diluted in MQ water to a final concentration of 25 M immediately prior to use. Analysis with Tris-Tricine gels indicate that the vast majority of A␤ in this preparation (referred to as sA␤) was in the monomer form (data not shown). To prepare A␤ fibrils (fA␤), 5 mM A␤ 1-42 or A␤ 1-40 in Me 2 SO was diluted in 10 mM HCl to 100 M (for A␤  or 200 M (for A␤ , vortexed for 30 s, and incubated at 37°C for 5 days (27). For mass spectrometry (MS) experiments, fA␤ was further purified by centrifugation at 14,000 ϫ g for 30 min. Fibrils were confirmed by thioflavin-T (ThT) fluorescence and electron microscopy (see below). Some experiments were repeated using A␤ 1-42 from a different manufacturer (American Peptide). ThT assay: 10 l of sample was added to 0.2 ml of 10 M ThT in 50 mM glycine buffer (pH 10.5). Fluorescence intensity was monitored at ex of 440 nm and em of 530 nm (28).
Proteolytic Digestions-Human recombinant ECE, ICE, NEP, and MMP-9 (R&D Systems) were used for digestion reactions using the following buffers: ECE, 0.1 M MES, 0.1 M NaCl (pH 6.0); IDE, 50 mM Tris, 1 M NaCl (pH 7.5); NEP, 0.1 M MES (pH 6.5); MMP-9, 50 mM Tris-HCl (pH 7.5), 10 mM CaCl 2 , 150 mM NaCl, 0.05% Brij 35. Pro-MMP-9 was activated with 1 mM p-aminophenylmercuric acetate at 37°C for 24 h prior to use. For sA␤ digestions, 100 nM protease was incubated with 2.5 M sA␤ at 37°C for varying periods of time (4 h to 5 days), then analyzed by MS. For fA␤ digestions (employing ThT or TEM), 200 nM protease was added to 10 l of fA␤ in reaction buffer and incubated at 37°C for 4 h to 5 days. After digestion, the reaction was analyzed by ThT assay or TEM. For fA␤ digestions (employing MS), 10 l of fA␤ was applied to a Microcon 50-kDa centrifugal ultrafiltration unit (YM-50, Millipore) and washed with 500 l of 10 mM HCl three times, then equilibrated with 500 l of MQ water. Thirty l of reaction buffer was added to the fA␤ (retentate) and centrifuged; the filtrate was analyzed by MS (see below) to ensure that any contaminating A␤ monomer or fragments were washed away. Ten l of 300 nM protease and 20 l of reaction buffer was added to the fA␤ (retentate) and incubated at 37°C for 24 h, then centrifuged and analyzed by MS to detect liberated fragments of fA␤.
Mass Spectrometry-Samples were first passed through a reverse phase C18 Ziptip (Millipore), according to the manufacturer's instructions, then diluted 1:1 with matrix solution (a saturated solution of 3,5-dimethoxy-4-hydroxycinnamic acid in 50% acetonitrile with 0.1% trifluoroacetic acid in water), loaded onto a plate, and allowed to dry. The sample was then analyzed on an Applied Biosystesms (Foster City, CA) Voyager DE-STR matrix-assisted laser desorption-ionization time-offlight mass spectrometer (MALDI-TOF MS) operated in linear mode. Insulin was used as an internal standard, and a calibration line constructed from its [M ϩ H] ϩ and [M ϩ 2H] 2ϩ ions was used for quantification (29).
Electron Microscopy-Small volumes (5 l) of fA␤ samples were absorbed onto glow-discharged, carbon-coated, formvar/ carbon-filmed 150-mesh copper grids. Grids were stained with 0.5% uranyl acetate and dried in a light-protected environment overnight before being viewed in a TEM (H-7500) operated at 80 kV (30).
In Situ Plaque Degradation-Brains were removed from anesthetized 9-month-old APP/PS1 mice after perfusion with cold saline and snap-frozen on dry ice. Five-m cryostat sections were collected on slides. Every other section was flipped 180°so that identical faces of adjacent sections were exposed. Paired adjacent sections (one incubated with buffer, the other with 70 nM protease) were incubated at 37°C for 5 days, stained with thioflavin-S (ThS) (31), then imaged with fluorescence microscopy (Olympus BX60). Using a 20ϫ objective lens, identical fields from paired sections were photographed (precise alignment was confirmed by superimposing plaque staining patterns). The area of ThS fluorescence was determined using image analysis software (Sigma Scan) and expressed as a fraction of total area. Fractional area was compared between paired sections. The specific activities (A␤ degrading activity) of all proteases were approximately equivalent (Fig. 2C).
In Situ Zymography-Fresh frozen sections (10 m) were incubated with DQ-gelatin (EnzCheck; Molecular Probes) at a concentration of 0.1 mg/ml in 1% low gel temperature agarose (Sigma) in PBS containing 4Ј,6-diamidino-2-phenylindole (1.0 g/ml). Sections were incubated for 24 h at room temperature. For further immunostaining, sections were washed and fixed with ice-cold acetone and alcohol (1:1), then subjected to immunofluorescent staining as described above. Specificity of MMP activity was confirmed by the addition of 20 mM EDTA, which inhibits activity (34).
Statistics-Student's t test was employed to determine differences between two groups (ThT comparison Ϯ protease). Paired comparisons (in situ plaque degradation assay) were analyzed using the paired t test. p values Ͻ0.05 were considered statistically significant.

RESULTS
Cleavage Sites for A␤-degrading Proteases-Of the candidate proteases capable of degrading sA␤ in vitro, only ECE, IDE, and NEP have evidence supporting a role in vivo (14 -16). To directly compare A␤-degrading activity and specific peptide fragments generated by digestion between these proteases and MMP-9, freshly prepared synthetic human A␤ 1-42 or A␤ 1-40 was incubated with each protease for 4 h prior to analysis with MALDI-TOF MS. Incubation of A␤ 1-42 with each recombinant human protease (ECE, IDE, NEP, and MMP-9) resulted in relatively specific profiles of A␤ fragments (Fig. 1, A-F ). The putative A␤ 1-42 cleavage sites for each protease, based on the fragments generated, are shown in Fig. 1G. The major fragments generated from proteolytic cleavage of A␤ 1-40 were similar to those generated from A␤ 1-42 (supplemental Fig. 1). In general, the A␤ cleavage sites are in good agreement with previous studies (10,13,24,35,36). However, a few differences were noted, possibly reflecting differences in the sources of the enzymes used (human versus rat), in the nature of A␤ substrate, in the digestion conditions (e.g. different buffers), or in detection and purification methods. MMP-9 generated six fragments that, in contrast to other proteases, were mostly in the hydrophobic C-terminal region of the A␤ peptide. All proteases examined exhibited a specific activity for sA␤ degradation that was remarkably similar (see below).
MMP-9 Degrades fA␤ in Vitro-To determine whether sA␤degrading proteases could degrade fA␤, human recombinant proteases (ECE, IDE, NEP, and MMP-9) or buffer alone were incubated with preformed A␤ fibrils. These fibrils were formed by incubating A␤ 1-42 or A␤ 1-40 at 37°C for 5 days (see "Experimental Procedures"). For A␤ 1-42 , ThT fluorescence reached maximal values within 1 day, while A␤ 1-40 reached maximal fluorescence within 3 days. Preformed A␤ fibrils were then incubated with proteases for 1-5 days at 37°C and examined using ThT fluorescence. Of the four proteases examined, only MMP-9 reduced ThT fluorescence compared with buffer controls (fA␤ 1-42 , Fig. 2A, fA␤ 1-40 ; supplemental Fig. 2), suggesting that fA␤ was degraded by this protease. This decrease in ThT fluorescence was time-dependent and declined to 70% of initial values by 96 h (Fig. 2B). To ensure that similar results could be obtained with A␤ preparations from different batches or manufacturers, we repeated the digestions using fA␤  from two different sources (American Peptide and Bachem) and at least two different lots and obtained similar results in each case (data not shown). Because our experiments directly compared the proteolytic activity of four different enzymes, we determined the specific activity for sA␤-degradation using MALDI-TOF MS data after 0 and 5 days of incubating proteases alone (without substrate) at 37°C. Specific sA␤ degrading activity was remarkably similar for all four proteases at 0 days, declining to ϳ50% after 5 days (Fig. 2C). Therefore, the observed differences in fA␤ degrading activity did not reflect different specific activities for sA␤ degradation among the proteases.
To visualize ultrastructural changes in A␤ fibrils after incubation with proteases, we performed TEM on preformed fA␤ incubated in buffer with or without the proteases. TEM of aggregated synthetic human A␤ 1-42 demonstrated fibrils with an approximate diameter of 10 nm, consistent with prior reports (37,38). In addition, thicker aggregates of fibers (80 -100 nm diameter) were occasionally observed. Incubation of fA␤ with ECE, IDE, or NEP, pro-MMP-9 produced no changes compared with PBS alone (Fig.  3, A-D), but MMP-9 incubation mixtures contained fewer fibrils. In addition, amorphous structures suggestive of decomposed  fibrils were observed in the MMP-9 incubation mixtures (Fig. 3, F  and G). These amorphous structures were very similar to that observed for ␣-helical structures (39) or after A␤ fibril deaggregation with curcumin (30).
To explore the potential mechanisms of fibril disruption by MMP-9, we determined whether A␤ fragments were released by MMP-9 digestion of fA␤. A␤ fragments generated by incubating purified fA␤ with ECE, IDE, NEP, or MMP-9 were isolated and analyzed by MALDI-TOF MS. MMP-9 was the only protease examined that produced A␤ fragments (Fig. 4C), with molecular masses of 2461.68 and 3390.59 daltons corresponding to A␤ 1-20 and A␤  . Inhibition of MMP-9 activity with EDTA prevented the generation of A␤ fragments (Fig. 4B), suggesting that proteolytic cleavage at Phe 20 -Ala 21 or Ala 30 -Ile 31 may be important for fibril degradation. Impurities in the MMP-9 preparation are indicated by * and are also seen in the MS of the protease alone (supplemental Fig. 3). Masses of the impurities did not coincide with A␤ fragments.
MMP-9 Degrades Compact Plaques in Situ-Although amyloid plaques are primarily composed of the A␤ peptide, numerous other components are known to colocalize in senile plaques, including proteoglycans, serum constituents, metal ions, and others. Furthermore, posttranslational modifications of A␤ may further stabilize amyloid deposits in plaques (40) Thus, the ability of an enzyme to degrade A␤ fibrils need not imply an ability to degrade amyloid plaques. To determine whether MMP-9 degrades amyloid plaques, we developed a sensitive method to compare the load of compact plaques in adjacent brain sections from aged APP/PS1 mice. Adjacent brain sections were incubated with proteases or buffer alone and then stained with ThS, and the area of fluorescence was compared between matching high power fields from adjacent sections. The area of ThS fluorescence (thioflavin load) in brain sections incubated with ECE, IDE, or NEP did not differ from those incubated with buffer alone, but incubation with MMP-9 resulted in a significant decrease in thioflavin load (Fig. 5A-D), suggesting that MMP-9 could degrade compact amyloid plaques. Fig. 5E illustrates a representative comparison between two corresponding high power fields from adjacent sections (upper panels) and the same fields with contrast thresholding to illustrate plaque load (lower panels). Note the relative decrease in plaque sizes in the section incubated with MMP-9 (arrows, ϩMMP-9) compared with that incubated with buffer alone (ϪMMP-9).

MMP-9 Expression and Activity in Brains of APPsw Mice-
Unlike other MMPs, MMP-9 is expressed in specific cells at low basal levels, and its expression can be induced by a variety of stimuli, including growth factors, cytokines, reactive oxygen species, or stressors (41). We examined MMP-9 expression in the brains of aged APPsw, APP/PS1, and wild-type littermate mice. Aged wild-type mice demonstrated a few isolated cells with MMP-9 immunoreactivity localized primarily in the corpus callosum (Fig. 6A), while APPsw mice had many more cells with prominent MMP-9 immunostaining throughout the brain (including cortex, corpus callosum, and hippocampus, Fig. 6, B and C). Double staining with ThS revealed that many of the MMP-9 immunoreactive cells appeared to surround ThS-positive compact plaques (Fig. 6C). Moreover, the cells had the appearance of activated astrocytes, based on their hypertrophic cell bodies and immunoreactivity with anti-GFAP antibodies (as illustrated by the double labeling experiment in Fig. 6, D-F). Similar expression profiles were found in 9-month-old APP/ PS1 mice (data not shown).
MMP-9 protein expression does not necessarily imply proteolytic activity. MMP-9 activity, like that of other MMPs, is stringently controlled through multiple mechanisms including transcriptional regulation, release from cells, activation through propeptide cleavage, and interaction with TIMPs. Thus, to examine net MMP-9 (gelatinase) activity in the brains of APPsw mice, in situ zymography was performed by incubating brain sections with fluorescently labeled gelatin. Gelatinase activity was detected within many but not all amyloid plaques (Fig. 6, G-I) in aged APPsw mice. Double labeling with ThS and anti-MMP-9 antibodies revealed that the vast majority of plaques with gelatinase activity were ThS-positive compact plaques surrounded by astrocytes expressing MMP-9 (data not shown). Gelatinase fluorescence was inhibited by EDTA even in areas of amyloid plaques (data not shown), suggesting that in situ zymography specifically reflects MMP activity.

DISCUSSION
This study provides evidence that MMP-9 is capable of degrading fibrillar A␤  , and this property is not shared by other sA␤-degrading proteases examined (including ECE, ICE, and NEP). MMP-9 incubated with fA␤ decreased ThT fluorescence compared with buffer controls. TEM examination of fibrils after digestion with MMP-9 revealed amorphous structures suggestive of disrupted fibrils. Furthermore, MALDI- Purified fA␤ 1-42 was washed clear of monomer A␤ and fragments using an ultrafiltration unit as described under "Experimental Procedures." After incubation with digestion buffer (A) or proteases (300 nM) for 24 h at 37°C, ultrafiltrate from the digestions were analyzed using MALDI-TOF MS. A␤ fragments were not detected after incubation with ECE, IDE, or NEP (data not shown); however, two fragments were detected after MMP-9 digestion and identified as A␤ fragments 1-20 and 1-30 (C ). These fragments were not detected in MMP-9 digests inhibited by EDTA (B). Four other peaks (labeled *) were identified as contaminants of the MMP-9 preparation and were present in MMP-9treated samples with (B) or without EDTA (C ), as well as in MMP-9 samples alone (data not shown).
TOF MS analyses of MMP-9 digestions demonstrated the appearance of two A␤ fragments (1-20 and 1-30) and established an association between A␤ digestion (at two cleavage sites characteristic of MMP-9) and the breakdown of fibrils. The process of fA␤ digestion appeared to be inefficient because ThT fluorescence declined slowly over days. We also demonstrate that MMP-9 is capable of degrading compact amyloid plaques in brain sections from aged APP/PS1 mice. Sensitive analyses of plaque load revealed smaller plaques in brain sections from aged APP/PS1 mice incubated with MMP-9 compared with adjacent section incubated with buffer alone. All of these activities were absent in parallel experiments using ECE, ICE, and NEP, despite the fact that all of the proteases had very similar specific activities for degradation of sA␤.
The fragments of A␤ that were released as fA␤ was degraded may provide some insight into the mechanism of fibril cleavage.
The precise structure of amyloid fibrils is not yet known, but several structural models have been proposed (42). Common to all of these models is a ␤-pleated sheet structure perpendicular to the fibril axis (43,44) with a hairpin loop at the C terminus of A␤ (45). Studies examining different fragments of A␤ demonstrate the importance of two critical regions for aggregation at approximately positions 17-21 and 30 -35 (46 -48). In our sA␤ degradation experiments, MMP-9 demonstrated cleavage sites mainly in the C-terminal hydrophobic region, which contrasted with the other proteases (see Fig. 1). Moreover, fA␤ degradation liberated A␤ fragments 1-20 and 1-30, and both cleavage sites are within sequences required for hairpin loop formation and ␤-sheet structure. Our results also suggest that the cleavage sites Phe 20 -Ala 21 and Ala 30 -Ile 31 are exposed on the surface of the fibrils in a manner that permits access (albeit inefficient) to MMP-9 for cleavage.
Fibrillar A␤ in compact plaques is believed to be extremely resistant to degradation and clearance, but growing evidence suggests that endogenous mechanisms for plaque clearance exist. The observation that the amyloid plaque size in brains of AD patients does not invariably increase with disease duration suggests an equilibrium between plaque formation and clearance (4,49). Furthermore, direct observations of amyloid plaques in vivo in aged APPsw mice demonstrate that some isolated plaques decrease in size over time (50), suggesting that there are endogenous mechanisms for plaque clearance. Consistent with this view, adult astrocytes incubated on amyloid-laden brain slices from aged PDAPP mice decreased the load of amyloid compared with control slices (51) via an ApoE-dependent mechanism (52). Astrocytes were found to be intimately associated with amyloid plaques in the in situ preparations (51) and in post-mortem brain tissue from AD cases (53). In the present study, we have shown that MMP-9 expression is increased in activated astrocytes surrounding diffuse and compact plaques and that gelatinase activity (which reflects MMP activity) resides almost exclusively in compact plaques. Furthermore, we have shown that MMP-9 can degrade compact plaques in an in situ degradation assay. These findings suggest that MMP-9 may contribute to astrocyte-mediated amyloid plaque degradation. FIGURE 5. In situ compact plaque degradation. Adjacent 5-m fresh frozen brain slices from aged APP/PS1 mice were incubated with the indicated proteases (70 nM) or buffer, respectively, at 37°C for 5 days. Sections were then fixed and stained with ThS, and plaque load was quantified and expressed as percent area per high power field (see "Experimental Procedures"). Incubation with MMP-9 (D), but not ECE (A), IDE (B), or NEP (C ), decreased the area of ThS staining compared with buffer controls (*, p Ͻ 0.05, paired t test). E, a representative example of ThS-stained compact plaques (arrows) in adjacent sections incubated with buffer alone (ϪMMP-9) or with activated MMP-9 (ϩMMP-9) is shown (upper panel). Plaque load was determined using image analysis software after a fixed adjustment of contrast threshold (Sigma Scan, lower panel). Bar, 100 m.
That MMP-9 is expressed in astrocytes surrounding amyloid plaques is consistent with previous reports of MMP-9 expression in neurofibrillary tangles, senile plaques, and in the vascular walls of postmortem brains from AD patients (13,24), and from APPsw mice (54). Furthermore, in vitro studies demonstrate that A␤ induces MMP-9 expression and activity in astrocytes (25,53). The precise role of activated astrocytes in the pathogenesis of AD is unclear. Some have suggested that astrocytes play a role in plaque formation and exacerbation of injury (53), but others suggest that astroctyes are involved in amyloid clearance (51). Activated astrocytes surrounding amyloid plaques have been shown to secrete a variety of pro-inflammatory molecules, such as interleukins, prostaglandins, leukotrienes, thromboxanes, and proteases (55). MMP-9 expression is frequently increased in pathological inflammatory responses, such as asthma (56), arthritis (57), and multiple sclerosis (58). The precise role of astrocytosis in promoting neurodegeneration versus amyloid clearance remains to be delineated.
If proteases such as MMP-9 play a role in degrading fA␤ and are expressed in activated astrocytes surrounding plaques, why do plaques continue to form in disease? One possible explanation is that fA␤-degrading proteases may be overwhelmed by the levels of fA␤ in the diseased brain. Based on the time course of fA␤ degradation in vitro, it appears that this activity of MMP-9 is slow and inefficient, and it might thus be readily overwhelmed. A second possibility is that proteases such as MMP-9 are induced after plaque formation and play a role in limiting plaque growth, causing plaques to reach an equilibrium at a given size. Another explanation may involve aberrant regulation of MMP-9 activity in disease. Like other MMPs, the net activity of MMP-9 depends on several factors, including expression level, release from cells, proteolytic activation, and interaction with endogenous inhibitors (TIMPs). Any of these regulatory processes could be altered in AD. One example is the expression of TIMP-1, the specific inhibitor of MMP-9, which has been reported to be increased in the CSF of AD patients (59).
In this study, we also demonstrate that MMP-9 is capable of degrading sA␤ with several cleavage sites toward the C terminus of the peptide. Therefore, like ECE, IDE, and NEP, MMP-9 may also play a role in the degradation and clearance of sA␤ in vivo. Indeed, we have evidence that steady-state levels of A␤ are lower in the brains of MMP-9 knock-out mice compared with wildtype controls. 4 In summary, we demonstrate that MMP-9 can degrade fA␤ and compact plaques and that this ability is not shared by other proteases examined here. Furthermore, MMP-9 is expressed in astrocytes surrounding plaques in the brains of aged APP/PS1 mice, and its activity is specifically detected in compact plaques. It is likely that other proteases may possess similar activity. Our findings contribute to accumulating evidence that endogenous mechanisms for clearing compact plaque exist. A better understanding of these mechanisms may identify potential therapeutic targets for this incurable disease.