Degradation of the Alzheimer Disease Amyloid β-Peptide by Metal-dependent Up-regulation of Metalloprotease Activity*

Biometals play an important role in Alzheimer disease, and recent reports have described the development of potential therapeutic agents based on modulation of metal bioavailability. The metal ligand clioquinol (CQ) has shown promising results in animal models and small phase clinical trials; however, the actual mode of action in vivo has not been determined. We now report a novel effect of CQ on amyloid β-peptide (Aβ) metabolism in cell culture. Treatment of Chinese hamster ovary cells overexpressing amyloid precursor protein with CQ and Cu2+ or Zn2+ resulted in an ∼85–90% reduction of secreted Aβ-(1–40) and Aβ-(1–42) compared with untreated controls. Analogous effects were seen in amyloid precursor protein-overexpressing neuroblastoma cells. The secreted Aβ was rapidly degraded through up-regulation of matrix metalloprotease (MMP)-2 and MMP-3 after addition of CQ and Cu2+. MMP activity was increased through activation of phosphoinositol 3-kinase and JNK. CQ and Cu2+ also promoted phosphorylation of glycogen synthase kinase-3, and this potentiated activation of JNK and loss of Aβ-(1–40). Our findings identify an alternative mechanism of action for CQ in the reduction of Aβ deposition in the brains of CQ-treated animals and potentially in Alzheimer disease patients.

Generation of APP-transfected CHO and N2a Neuroblastoma Cells-APP-CHO and APP-N2a neuroblastoma cells were generated by expressing the 695-amino acid APP cDNA in the pIRESpuro2 expression vector (Clontech). Cells were transfected using Lipofectamine 2000 and cultured in RPMI 1640 medium supplemented with 1 mM glutamine and 10% fetal bovine serum (all from Invitrogen, Mount Waverley, Victoria). Transfected cells were selected and maintained using 7.5 g/ml puromycin (Sigma).
Exposure of Cells to CQ and Metals-APP-overexpressing cells were passaged at a ratio of 1:6 and grown in 6-or 12-well plates for 2-3 days before experiments. CQ was prepared as a 10 mM stock solution in Me 2 SO and added to serum-free RPMI 1640 medium supplemented with puromycin as described above. Basal metal levels in the medium were 0.5, 1.3, and 2.1 M for Cu 2ϩ , Zn 2ϩ , and Fe 2ϩ , respectively, as determined by inductively coupled plasma mass spectrometry (ICP-MS). Additional metals were added (10 M unless stated otherwise), and the medium was briefly mixed by aspiration prior to addition to cells. Control cultures were treated with vehicle (Me 2 SO) alone. Inhibitors of phosphoinositol 3-kinase (PI3K) (LY-294,002 and wortmannin), JNK (SP600125), MEK1/2 (PD 98,059), p38 (SB 203580), GSK3 (GSK Inhibitor IX), and metalloproteases (GM 6001, phosphoramidon, thiorphan, bestatin, MMP-2 Inhibitor I, and MMP-9 Inhibitor I) were prepared as 10 mM stock solutions in Me 2 SO and added at the indicated concentrations. Ascorbate, MnTMPyP, bacitracin, BPS, LiCl, and MMP-3 Inhibitor I were prepared as 10 mM solutions in distilled H 2 O. Serine/cysteine protease inhibitor mixture (EDTA-free) was prepared as a 10ϫ solution in distilled H 2 O. Where stated, vector only-transfected or wild-type (non-APP-overexpressing) cells were exposed to synthetic human A␤-(1-40) with or without CQ, metals, and inhibitors (see below). Cultures were incubated for up to 6 h, and conditioned media were taken for measurement of A␤ levels by enzyme-linked immunosorbent assay (ELISA). Cell viability was determined by lactate dehydrogenase release following kit instructions (Promega Corp., Annandale, New South Wales, Australia). For immunoblotting, cells were harvested into Phos-phoSafe extraction buffer (Novagen) containing Protease Inhibitor Cocktail III (Calbiochem) and stored at Ϫ80°C until used. Alternatively, cells were washed three times with phosphate-buffered saline (PBS) and harvested for analysis of metal levels by ICP-MS.
ICP-MS-Cells were treated with CQ and/or metals for 6 h unless stated otherwise and washed three times with Chelex 100-treated PBS (pH 7.4). Cells were scraped into PBS; an aliquot was taken for protein determination (protein microassay, Bio-Rad); and the remaining cells were collected by centrifugation at 14,000 rpm for 2 min in a Hermle microcentrifuge (Labnet International, Inc., Edison, NJ). Metal levels were determined in cell pellets by ICP-MS as described previously (12) and converted to ng of metal/mg of protein.
Degradation of Synthetic A␤- -Human A␤-(1-40) was purchased from the W. M. Keck Laboratory (Yale University, New Haven, CT) and dissolved in Me 2 SO at 1 mg/ml. The dissolved peptide was further diluted into Chelex 100-treated distilled H 2 O at 100 ng/ml before addition to vector only-transfected CHO cell cultures in serum-free medium at 10 ng/ml without aging. In separate experiments, A␤-  was also added to N2a mouse neuroblastoma, SH-SY5Y human neuroblastoma, or HeLa human epithelial cells in serum-free Opti-MEM I (Invitrogen). After 6 h (with or without addition of inhibitors and 10 M each CQ, Cu 2ϩ , or CQ and Cu 2ϩ ), the medium was collected, and the remaining A␤-(1-40) levels were determined by ELISA.
Double Antibody Capture ELISA for A␤ Detection-A␤ levels were determined in culture medium using the DELFIA double capture ELISA (PerkinElmer Life Sciences, Melbourne, Australia). 384-Well plates (Greiner Bio-One GmbH, Frickenhausen, Germany) were coated with monoclonal antibody G210 in 15 mM Na 2 CO 3 and 35 mM NaHCO 3 (pH 9.6) for A␤-(1-40) detection. Plates were washed with PBS containing 0.05% Tween and blocked with 0.5% (w/v) casein. Biotinylated monoclonal antibody WO2 (A␤-(5-8) epitope) and the culture medium or A␤ standards were added (50 l) to each well and incubated overnight at 4°C. Plates were washed with PBS containing 0.05% Tween, and streptavidin-labeled europium (PerkinElmer Life Sciences) was added. The plates were washed; enhancement solution (PerkinElmer Life Sciences) was added; and the plates were read in a Wallac VICTOR 2 plate reader with excitation at 340 nM and emission at 613 nM. A␤-(1-40) and A␤-(1-42) standards and samples were assayed in triplicate. The values obtained from the triplicate wells were used to calculate the A␤ concentration (expressed as ng/ml) based on the standard curve generated on each plate. We observed a good correlation between ELISA results and Western blot analysis of A␤ levels in CQ⅐Cu 2ϩ -treated cultures. As the ELISA offered quantitative data on A␤ levels, we chose this as the preferred method for assessing changes to secreted A␤ levels.
Western Blot Analysis of Protein Expression and Phosphorylation-Cell lysates prepared in PhosphoSafe extraction buffer were mixed with SDS sample buffer (Novex) and separated on 12% Tris/glycine/SDSpolyacrylamide gels (Novex). Western blotting of A␤ in the conditioned medium was performed using 10 -20% Tris/Tricine gels. Proteins were transferred to polyvinylidene difluoride membranes and blocked with milk solution in Tris-buffered saline/Tween before immunoblotting for total or phospho-specific proteins. Membranes were probed for 1 h with antiserum against A␤ (antibody WO2), C-terminal APP (antibody 369), or full-length APP (antibody 22C11) at 1:2000 dilution and with horseradish peroxidase-conjugated rabbit anti-mouse or goat anti-rabbit secondary antibody at 1:5000 dilution. For detection of signal transduction molecules, membranes were probed with polyclonal antiserum against actin, JNK, phospho-JNK, p38, phospho-p38, ERK1/2, phospho-ERK1/2, Akt, phospho-Akt, GSK3␤, phospho-GSK3␣/␤, MMP-2, or MMP-6 at 1:5000 dilution. Horseradish peroxidase-conjugated goat anti-rabbit secondary antiserum was used at 1:10,000 dilution. Blots were developed by chemiluminescence (ECL Advance, Amersham Biosciences) and imaged on a GeneGnome chemiluminescence imager (Syngene, Cambridge, UK). We found that the expression of total levels of kinases (Akt, JNK, ERK, and p38) was unaffected by metal uptake in APP-CHO cells. In contrast, actin, tubulin, and other proteins normally used for equalizing protein loading were found to be altered depending on metal levels. 5 Therefore, equal sample loading and protein transfer were assessed by consistency of total kinase protein levels rather than unrelated protein levels on immunoblots.
Cell Adhesion Assay-Cell adhesion to collagen type IV was determined using an InnoCyte ECM cell adhesion assay (collagen type IV; Merck Biosciences). Cells were treated with CQ, Cu 2ϩ , or CQ and Cu 2ϩ (with or without inhibitors) for 4 h before harvesting with a rubber policeman into the culture medium. Cells were dissociated by aspiration, replated onto collagen type IV, and cultured for an additional 2 h. The medium was discarded, and cells were washed briefly with two changes of PBS before addition of calcein acetoxymethyl ester for 1 h (37°C). Cell adhesion was determined by fluorescence spectrophotometry on a Wallac VICTOR 2 plate reader with excitation at 490 nM and emission at 535 nm.
Statistical Analysis-All data described in graphical representations are means Ϯ S.E. unless stated from a minimum of three separate experiments. Results were analyzed using Student's two-tailed t test.

RESULTS
CQ Mediates Uptake of Cu 2ϩ and Zn 2ϩ but Not Fe 2ϩ in APP-CHO Cells-As CQ is a lipid-soluble metal ligand, we examined the effect of CQ on metal levels in APP695-transfected CHO cells (APP-CHO). Cultures were treated with CQ (10 M) alone or in the presence of 10 M Cu 2ϩ , Zn 2ϩ , or Fe 2ϩ for 6 h, and cellular metal levels were assessed by ICP-MS. Basal Cu 2ϩ levels were 4.6 Ϯ 0.1 ng/mg of protein, and exposure to CQ alone increased this to 20 Ϯ 3 ng/mg of protein ( p Ͻ 0.01) ( Table 1). Treatment with CQ and Cu 2ϩ (CQ⅐Cu 2ϩ ) induced a dramatic 103-fold increase in cellular Cu 2ϩ levels (472 Ϯ 46 ng/mg of protein; p Ͻ 0.005) ( Table 1). CQ also increased cellular Zn 2ϩ levels from 182 Ϯ 8 to 1838 Ϯ 64 ng/mg of protein ( p Ͻ 0.001) ( Table 1). Measurement of cell survival (lactate dehydrogenase release) revealed no significant effect on cell viability after 6 h of exposure to 10 M CQ and Cu 2ϩ or Zn 2ϩ . Treatment of cultures with Fe 2ϩ alone (10 M) resulted in a 16-fold FIGURE 1. A, A␤-(1-40) levels in medium from CQtreated APP-CHO cells. Cultures were exposed to CQ (10 M) with or without 10 M Cu 2ϩ , Zn 2ϩ , or Fe 2ϩ for 6 h, and A␤-(1-40) levels were determined in culture medium by ELISA. Cu 2ϩ alone induced a small but significant increase (*, p Ͻ 0.01) in A␤-(1-40) secretion, whereas exposure to CQ⅐Cu 2ϩ or CQ⅐Zn 2ϩ significantly reduced A␤-(1-40) levels (**, p Ͻ 0.0001). B, A␤-(1-42) levels in medium from CQ-treated APP-CHO cells. Cultures were exposed to 10 M CQ with or without 10 M Cu 2ϩ as described for A. CQ⅐Cu 2ϩ induced a significant decrease in secreted A␤-(1-42) levels (**, p Ͻ 0.0001). Error bars represent S.E. C, immunoblotting of medium from CQ⅐Cu 2ϩtreated APP-CHO cells. Cultures were exposed to 10 M CQ⅐Cu 2ϩ for 6 h, and the conditioned medium was analyzed by Western blotting using anti-A␤ antiserum (antibody WO2). A␤ levels were significantly lower in cultures treated with CQ⅐Cu 2ϩ compared with untreated controls.  (Fig. 1A). Interestingly, 10 M Cu 2ϩ alone for 6 h induced a 35% increase in A␤-(1-40) levels ( p Ͻ 0.01) (Fig. 1A).
When cultures were exposed to 10 M CQ and 10 M Cu 2ϩ , we observed a potent reduction (ϳ85%) of secreted A␤-(1-40) levels ( p Ͻ 0.0001) (Fig. 1A). An analogous effect was observed upon treatment with CQ and 10 M Zn 2ϩ (Fig. 1A). No significant changes were observed in A␤-(1-40) levels when cells were treated with CQ plus Fe 2ϩ (Fig. 1A). Potent inhibition of secreted A␤-(1-42) levels also occurred with CQ⅐Cu 2ϩ -treated cells (Fig. 1B). However, as A␤-(1-42) levels in APP-CHO cells were near the detection limit of the ELISA, subsequent analysis of A␤ was restricted to A␤-(1-40). The loss of secreted A␤ upon treatment with CQ⅐Cu 2ϩ was confirmed by immunoblot analysis of the conditioned medium ( Fig. 1C) and surface-enhanced laser desorption ionization mass spectrometry (data not shown).
Inhibition of A␤ Can Be Induced by Low Concentrations of CQ and Cu 2ϩ -To examine the potency of CQ in inhibiting secreted A␤ levels, we treated cultures with 0.1-50 M CQ with or without 10 M Cu 2ϩ for 6 h. A␤-(1-40) was significantly decreased at 0.1 and 1.0 M CQ plus Cu 2ϩ ( Fig.  2A). We also examined the effects of different concentrations of CQ on Cu 2ϩ uptake in APP-CHO cells. 0.1 M CQ induced an increase of ϳ25fold in cellular Cu 2ϩ levels (Fig. 2B). Increasing CQ concentrations resulted in further elevation of cellular Cu 2ϩ levels, reaching 112-fold (at 50 M) compared with control levels (Fig. 2B). The ability of low concentrations of CQ to increase cellular Cu 2ϩ levels correlated with the potent reduction of secreted A␤ levels by CQ⅐Cu 2ϩ (Fig. 2A). Although CQ has been reported FIGURE 2. A, A␤-(1-40) levels in medium from APP-CHO cells exposed to increasing concentrations of CQ. Cells were exposed to 0.1-50 M CQ with or without 10 M Cu 2ϩ for 6 h. A␤-(1-40) levels were determined in the culture medium by ELISA. All concentrations of CQ⅐Cu 2ϩ tested (0.1-50 M) induced a significant loss of secreted A␤-(1-40) compared with CQ alone ( p Ͻ 0.001-0.0001). B, cellular Cu 2ϩ levels in APP-CHO cells exposed to Cu 2ϩ and increasing concentrations of CQ. Cells were exposed to 10 M Cu 2ϩ and 0.1-50 M CQ for 6 h. Cu 2ϩ levels were determined in cell pellets by ICP-MS, revealing significantly increased cellular Cu 2ϩ levels at all concentrations of CQ ( p Ͻ 0.001-0.0001). The Cu 2ϩ levels in vehicletreated controls were equivalent to 1-fold Cu 2ϩ . C, A␤-(1-40) levels in medium from APP-CHO cells exposed to CQ and increasing concentrations of Cu 2ϩ . Cells were treated with 10 M CQ and 0.1-10 M Cu 2ϩ for 6 h. Cu 2ϩ alone induced a small increase in secreted A␤-(1-40) levels, but CQ⅐Cu 2ϩ induced a dose-dependent decrease in secreted A␤-(1-40) levels ( p Ͻ 0.01-0.0001). D, A␤-(1-40) levels in medium from APP-CHO cells exposed to CQ⅐Cu 2ϩ for different time periods. Cells were treated with CQ⅐Cu 2ϩ (10 M) as described above, and A␤-(1-40) levels were determined in the culture medium at time points up to 6 h after the start of treatment. Exposure to CQ⅐Cu 2ϩ induced a time-dependent decrease in A␤-(1-40) levels in the medium ( p Ͻ 0.05-0.0001). E, cellular Cu 2ϩ levels in APP-CHO cells exposed to CQ⅐Cu 2ϩ for different time periods. Cells were exposed to CQ⅐Cu 2ϩ (10 M), and cellular Cu 2ϩ levels were determined by ICP-MS in pellets at different time points up to 6 h after the start of treatment. CQ⅐Cu 2ϩ induced a time-dependent increase in cellular Cu 2ϩ levels ( p Ͻ 0.05 at 120 min and p Ͻ 0.001 at 360 min). Relatively little change in cellular Cu 2ϩ levels was induced by CQ or Cu 2ϩ alone. The Cu 2ϩ levels in vehicle-treated controls were equivalent to 1-fold Cu 2ϩ . For all graphs, error bars represent S.E. F, APP levels in APP-CHO cells treated with CQ⅐Cu 2ϩ . Cells were treated with 10 M CQ with or without 10 M Cu 2ϩ for 6 h, and APP expression was determined by Western blotting in cell lysates and the conditioned medium. Equal protein loading was confirmed by immunoblotting for total JNK (not shown). CQ alone or CQ⅐Cu 2ϩ decreased cellular APP (cAPP) expression and secreted APP (sAPP) levels in the conditioned medium. No change in APP ␤-C-terminal fragment (␤CTF) C99 was observed with any treatment, whereas CQ⅐Cu 2ϩ reduced expression of APP ␣-C-terminal fragment (␣CTF) C83. Changes in APP expression did not correlate with secreted A␤-(1-40) levels.
to optimally bind Cu 2ϩ at a ratio of 2:1 (13), our titration studies showed no significant differences in Cu 2ϩ uptake and inhibition of A␤ levels upon varying the CQ/Cu 2ϩ ratios.
To determine the effects of Cu 2ϩ concentration on secreted A␤ levels, cultures were exposed to 10 M CQ with different concentrations of Cu 2ϩ . 0.1 M added Cu 2ϩ significantly inhibited A␤ levels (Fig. 2C). Higher concentrations of added Cu 2ϩ further decreased secreted A␤ levels (Fig. 2C). We also examined the time course of A␤ inhibition by CQ plus Cu 2ϩ (10 M each). We observed an initial decrease in A␤ levels from 30 to 60 min after addition of CQ⅐Cu 2ϩ . A greater loss of A␤ was observed from 60 to 120 min after treatment (Fig. 2D). Examination of cellular metal levels revealed a 22-fold increase in Cu 2ϩ after a 10-min exposure to CQ⅐Cu 2ϩ (Fig. 2E). Cu 2ϩ levels increased further at each time point, reaching a maximum level of 103-fold at 360 min (Fig. 2E).

Loss of A␤ by CQ⅐Cu 2ϩ Does Not Correlate with Cellular APP Levels-
To further understand how CQ⅐Cu 2ϩ mediates A␤ loss, we determined whether there is a corresponding loss in APP expression. Exposure to CQ alone or to CQ⅐Cu 2ϩ reduced both APP expression and secretion (Fig. 2F ). However, as shown in Fig. 1A, only CQ⅐Cu 2ϩ inhibited secreted A␤ levels. Interestingly, there was a reduction in the ␣-Cterminal 83-amino acid fragment of APP (C83) upon CQ⅐Cu 2ϩ treatment, although no changes in APP ␤-C-terminal 99-amino acid fragment (C99) expression were found (Fig. 2F). This was consistent with our observation that the activity of BACE1 (beta-site APP-cleaving enzyme 1) in APP-CHO membrane preparations was unchanged after treatment with CQ⅐Cu 2ϩ . Likewise, analysis of COS-7 cells transfected with a C-terminal APP construct (APP C99) (14) revealed no effect on ␥-secretase cleavage of APP C99 by CQ⅐Cu 2ϩ . 5 These findings demon- FIGURE 3. A, effects of CQ and metals on activation of MAPK pathways in APP-CHO cells. Cells were exposed to 10 M CQ with or without 10 M Cu 2ϩ or Zn 2ϩ for 6 h. Activation of MAPKs, including JNK, p38, and ERK1/2, was determined by Western blotting of cell lysates. CQ alone had no effect on the activity of MAPKs. In the presence of Cu 2ϩ or Zn 2ϩ , CQ induced activation of JNK, p38, and ERK1/2 (phospho-JNK, phospho-p38, and phospho-ERK1/2) B, JNK activation is associated with the loss of A␤-(1-40) in APP-CHO cells treated with CQ⅐Cu 2ϩ . Cells were exposed to 10 M CQ⅐Cu 2ϩ for 6 h in the presence or absence of the JNK inhibitor SP600125 strate that the loss of secreted A␤ upon treatment with CQ⅐Cu 2ϩ is unlikely to result from altered APP processing.
Loss of Secreted A␤ by CQ⅐Cu 2ϩ Is Mediated through Activation of JNK and ERK-Metal ligands can stimulate MAPK pathways (15,16). To examine whether the effects of CQ⅐Cu 2ϩ on A␤ occur via these pathways, we treated cultures with CQ and Cu 2ϩ or Zn 2ϩ (10 M each) and measured activation of JNK, p38, and ERK1/2 in cell lysates. CQ with Cu 2ϩ or Zn 2ϩ induced substantial activation of JNK and ERK1/2, with moderate activation of p38 (Fig. 3A).
We then examined whether activation of these MAPK pathways is involved in the inhibitory action of CQ and metals on secreted A␤ levels. The JNK inhibitor SP600125 resulted in significant inhibition of JNK phosphorylation (Fig. 3B) and a significant elevation of A␤-(1-40) levels compared with CQ⅐Cu 2ϩ alone ( p Ͻ 0.001) (Fig. 3B). The ERK1/2 phosphorylation inhibitor PD 98,059 (5 M) prevented ERK activation after exposure to CQ⅐Cu 2ϩ (Fig. 3C) and significantly inhibited A␤ loss ( p Ͻ 0.001) (Fig. 3C). In contrast, the p38 inhibitor SB 203580 or the broad-spectrum protein kinase inhibitor staurosporine had no restorative effect on A␤ levels (Fig. 3, D and E).
JNK can be activated in response to cell stresses such as generation of reactive oxygen species or through growth factor-mediated pathways (17). Therefore, we examined whether JNK phosphorylation is mediated by generation of reactive oxygen species in the CQ⅐Cu 2ϩtreated cultures. APP-CHO cells were exposed to CQ⅐Cu 2ϩ together with the reactive oxygen species scavenger MnTMPyP (200 M) or the antioxidant ascorbate (1 mM). Treatment with these antioxidants did not inhibit JNK phosphorylation or prevent A␤ loss in CQ⅐Cu 2ϩtreated cultures (Fig. 3F). This is consistent with Zn 2ϩ inducing effects analogous to those of Cu 2ϩ , as Zn 2ϩ is a redox-inactive metal and should not directly stimulate reactive oxygen species generation. Therefore, the results strongly suggest that activation of JNK by CQ⅐Cu 2ϩ is not mediated through metal-induced oxidative stress.
Inhibition of A␤ by CQ⅐Cu 2ϩ Requires Activation of the PI3K-Akt-GSK3 Pathway-Modulation of GSK3, a downstream target of PI3K and Akt activation, changes A␤ production in APP-CHO cells (18). Therefore, we examined whether the PI3K-Akt-GSK3 pathway is associated with the loss of A␤ production in CQ⅐Cu 2ϩ -treated cells. Treatment of cells with CQ and Cu 2ϩ (10 M each) for 6 h resulted in significant activation of Akt (Fig. 4A). Co-treatment of cultures with the specific PI3K inhibitor LY-294,002 inhibited Akt phosphorylation induced by CQ⅐Cu 2ϩ and significantly abrogated the decrease in secreted A␤ levels ( p Ͻ 0.0001) (Fig. 4A).
Treatment of cultures with 10 M CQ⅐Cu 2ϩ for 6 h increased the phosphorylated forms of GSK3␣/␤, and this effect was blocked by LY-294,002 (Fig. 4A). There was also a small increase in total GSK3␤ levels in CQ⅐Cu 2ϩ -treated cultures, which may partially account for the increased levels of phosphorylated GSK3.
We then investigated whether PI3K-Akt-GSK3 activation is upstream of MAPK activation. Treatment of cultures with 25 M LY-294,002 (or 10 nM wortmannin; data not shown) inhibited phosphorylation of Akt as well as phosphorylation of both JNK and ERK1/2 (Fig. 4B). Conversely, treatment of cultures with inhibitors of JNK and ERK1/2 phosphorylation (SP600125 and PD 98,059 respectively) did not inhibit Akt phosphorylation (data not shown). These data demonstrate that PI3K-Akt activation is upstream of JNK and ERK activation.
Activation of the PI3K-Akt and JNK Pathways Alone Is Not Sufficient for Loss of A␤-As inhibitors of PI3K and JNK pathways blocked the loss of A␤ by CQ⅐Cu 2ϩ , we examined whether nonspecific up-regulation of these pathways also results in loss of A␤ in APP-CHO cells. Cultures exposed to 25-100 M Cu 2ϩ (without CQ) for 6 h revealed potent activation of Akt, whereas 50 and 100 M Cu 2ϩ also induced phosphorylation of GSK3 and JNK (Fig. 4C). Moreover, cultures treated with the apoptotic agent cisplatin (200 M) for 6 h revealed activation of JNK but not Akt (Fig. 4C). However, neither of these treatments (Cu 2ϩ or cisplatin) reduced secreted A␤ levels, demonstrating that the PI3K-Akt and JNK pathways are necessary, but insufficient alone, for the loss of A␤ in APP-CHO cells.
GSK3 Phosphorylation Promotes Activation of JNK in Cultures Treated with CQ⅐Cu 2ϩ -Our data suggested that phosphorylation of GSK3 in CQ⅐Cu 2ϩ -treated cells may modulate downstream JNK activation. To examine this, we treated cultures with CQ⅐Cu 2ϩ in the presence of LiCl (an inducer of GSK3 phosphorylation). In the presence of CQ⅐Cu 2ϩ , 5 mM LiCl increased phosphorylated GSK3 levels compared with CQ⅐Cu 2ϩ alone (Fig. 5A). LiCl had no effect on phospho-Akt levels, demonstrating that the effect was not mediated through increased PI3K and Akt activities (Fig. 5A). LiCl potentiated JNK phosphorylation in cultures treated with CQ⅐Cu 2ϩ (Fig. 5B). This potentiation was sufficient to overcome the inhibitory action of 25 or 50 M SP600125 on JNK phosphorylation (Fig. 5B). Interestingly, potentiation of JNK phosphorylation by LiCl also overcame the ability of SP600125 to prevent A␤ loss in the medium (Fig. 5B). To confirm the potentiating effect of GSK3 phosphorylation on JNK activation, we treated cultures with 1 M CQ and 10 M Cu 2ϩ in the presence of GSK Inhibitor IX (10 or 25 M). This increased phosphorylation of GSK3 and JNK compared with CQ⅐Cu 2ϩ alone (Fig. 5C). The increased GSK3 and JNK phosphorylation correlated with a down-regulation of A␤ levels in the culture medium (Fig.  5C). These results provide strong evidence that increased phosphorylation of GSK3 in CQ⅐Cu 2ϩ -treated cultures promotes activation of JNK and leads to loss of secreted A␤.
CQ⅐Cu 2ϩ Induces Metalloprotease-dependent Loss of A␤-Exposure of APP-CHO cells to CQ⅐Cu 2ϩ for 6 h resulted in morphological changes consistent with altered cell adhesion (detachment of cells), but without an obvious role for cytotoxicity or oxidative stress (Fig.  3F). To examine this, we measured adhesion of cells to a collagen type IV matrix after treatment with 10 M CQ and Cu 2ϩ . As shown in Fig. 6A, CQ⅐Cu 2ϩ inhibited APP-CHO cell adhesion to collagen type IV by ϳ50%. The loss of adhesion could be prevented by treatment with SP600125 or LY-294,002 (Fig. 6A). As loss of cell adhesion is commonly associated with activation of metalloproteases (19), we treated cells with broad-spectrum metalloprotease inhibitors. GM 6001 (10 M), BPS (500 M), and MMP Inhibitor I (20 M) significantly inhibited CQ⅐Cu 2ϩ -mediated loss of cell adhesion to collagen type IV (Fig. 6A).
To determine whether metalloproteases mediate A␤ loss, APP-CHO cells were treated with a range of metalloprotease inhibitors, and A␤ levels were measured after exposure to CQ⅐Cu 2ϩ . All metalloprotease inhibitors tested except thiorphan (neprilysin inhibitor) significantly inhibited the decrease in secreted A␤ levels induced by CQ⅐Cu 2ϩ (Fig. 6B). To confirm that the loss of secreted A␤ was mediated through increased metalloprotease-mediated degradation rather than altered APP processing, vector only-transfected CHO cell cultures were exposed to 10 ng/ml synthetic human A␤-(1-40) for 6 h with or without CQ⅐Cu 2ϩ . Measurement of A␤-(1-40) levels in the conditioned medium revealed 0.89 Ϯ 0.11 ng/ml remaining in the control medium after 6 h, indicating substantial clearance by cell uptake and/or degradation (Fig. 6C). Exposure of cultures to 10 M CQ or 10 M Cu 2ϩ increased the levels of synthetic A␤-  remaining in the medium after 6 h (Fig. 6C). However, treatment of cultures with CQ⅐Cu 2ϩ significantly decreased A␤-(1-40) levels by 56% compared with controls and by 77% compared with CQ alone (Fig. 6C). Interestingly, this effect was prevented by co-treatment of cultures with LY-294,002 (25 M) or inhibitors of metalloproteases (Fig. 6C). The results clearly support a role for PI3K-mediated metalloprotease degradation of A␤ as the primary cause of A␤ loss in cultures treated with CQ⅐Cu 2ϩ .
CQ⅐Cu 2ϩ Induces Up-regulation of MMP-2 and MMP-3 through Activation of the PI3K and JNK Pathways-The efficacy of GM 6001 and MMP Inhibitor I against loss of A␤ and cell adhesion strongly supported a role for up-regulation of MMPs in CQ⅐Cu 2ϩ -treated cultures. Therefore, we measured the activity of MMPs in cells treated with CQ⅐Cu 2ϩ using MMP-specific fluorescent substrates. MMP assays of cell lysates or the conditioned medium after treatment with CQ⅐Cu 2ϩ for 6 h revealed a significant elevation of the specific activities of MMP-2 and MMP-3 (Fig. 7A). No significant changes were observed in the activities of MMP-1, MMP-8, and MMP-9. Western blot analysis of cell lysates with antisera to MMP-2 and MMP-9 confirmed the results from the fluorescent substrate assay. Both latent and activated forms of MMP-2 were up-regulated in cultures exposed to CQ⅐Cu 2ϩ , whereas MMP-9 revealed only a minimum change (Fig. 7A, inset). FIGURE 5. A, effect of LiCl on GSK3 phosphorylation in APP-CHO cells exposed to CQ⅐Cu 2ϩ . Cells were treated with 10 M CQ⅐Cu 2ϩ with and without 5 mM LiCl for 6 h. Phosphorylation of GSK3␣/␤ was determined by Western blotting of cell lysates. LiCl alone induced a low level of GSK3 phosphorylation, but, together with CQ⅐Cu 2ϩ , greatly increased GSK3 phosphorylation without effect on Akt phosphorylation. B, effect of LiCl on A␤-  in cultures treated with CQ⅐Cu 2ϩ . APP-CHO cells were treated with 10 M CQ⅐Cu 2ϩ with or without the JNK inhibitor SP600125 (5, 25, or 50 M) and/or LiCl (5 mM) for 6 h. CQ⅐Cu 2ϩ induced JNK phosphorylation, and this was inhibited by co-treatment with SP600125. The loss of A␤-(1-40) induced by CQ⅐Cu 2ϩ was also inhibited by SP600125. Co-treatment with 5 mM LiCl substantially increased JNK activation even in the presence of the JNK inhibitor SP600125. Co-treatment with LiCl also restored the loss of A␤-   To further confirm activation of MMP-2 and MMP-3 by CQ⅐Cu 2ϩ , cultures were treated with selective MMP inhibitors. Incubation of cultures with MMP-2 Inhibitor I prevented activation of MMP-2 by CQ⅐Cu 2ϩ , but had no significant effect on MMP-3 activity (Fig. 7B). Likewise, MMP-3 Inhibitor I blocked activation of MMP-3, but did not affect MMP-2 activation by CQ⅐Cu 2ϩ (Fig. 7B). An MMP-9 inhibitor had no effect on either MMP-2 or MMP-3 activation by CQ⅐Cu 2ϩ (Fig.  7B). Moreover, we observed that LY-294,002 and SP600125 blocked activation of MMP-2 and MMP-3 (Fig. 7B).
Interestingly, the inhibitors of MMP-2 and MMP-3 significantly abrogated the loss of A␤-(1-40) caused by CQ⅐Cu 2ϩ (Fig. 7C). These effects were consistent with a previous report that both MMP-2 and MMP-3 can cleave A␤ at several sites (10). In fact, surface-enhanced laser desorption ionization analysis of medium from our control cultures revealed A␤ cleavage products consistent with MMP-2-mediated degradation. 5 Unfortunately, few A␤ fragments of any size were observed in CQ⅐Cu 2ϩ -treated cultures. This suggested that further degradation of the A␤ cleavage fragments may be occurring in these cultures, possibly through activation of aminopeptidases. This was supported by inhibition of A␤ loss by co-treatment with bestatin (aminopeptidase inhibitor) (Fig. 6B).
Finally, we examined alternative cell types for their ability to degrade A␤ when exposed to CQ⅐Cu 2ϩ . Treatment of APP-overexpressing N2a murine neuroblastoma cells with 10 M CQ and 10 M Cu 2ϩ for 6 h reduced A␤ levels from ϳ1.1 to 0.4 ng/ml ( p Ͻ 0.001), whereas CQ or Cu 2ϩ alone had no significant effect (Fig. 8A). In addition, non-transfected N2a, SH-SY5Y human neuroblastoma, and HeLa human epithelial cells were treated with 10 M CQ and Cu 2ϩ together with 10 ng/ml synthetic human A␤-(1-40). Measurement of A␤ levels in the conditioned medium after 6 h revealed significantly reduced synthetic A␤-(1-40) levels in all cell types after treatment with CQ⅐Cu 2ϩ , and this could be prevented by co-treatment with GM 6001 (Fig. 8B). These results demonstrate that CQ⅐Cu 2ϩ can modulate secreted A␤ levels via metalloproteases in different cell types, including neuroblastoma cells.

DISCUSSION
In this study, we have shown for the first time that the lipid-soluble metal ligand CQ modulates secreted A␤ levels in vitro. Whereas treatment of APP-expressing cells with CQ alone had little effect on A␤ levels in the culture medium, treatment with CQ complexed to Cu 2ϩ or Zn 2ϩ dramatically decreased extracellular A␤ levels. We have shown that this effect is closely related to the ability of CQ to mediate substantial increases in cellular Cu 2ϩ or Zn 2ϩ levels, resulting in selective upregulation of MMP activity.
Interestingly, we found that even low concentrations of CQ or Cu 2ϩ (0.1-1 M each) could induce a significant loss of A␤ after only 6 h. The potency with which CQ⅐Cu 2ϩ inhibited A␤ underscores the potential physiological relevance of our findings. A recent study reported human plasma levels of CQ at ϳ13-25 M during small FIGURE 6. A, cell adhesion to a collagen type IV matrix. APP-CHO cells were exposed to CQ (10 M), Cu 2ϩ (10 M), or CQ⅐Cu 2ϩ with or without inhibitors, and adhesion to collagen type IV was determined by uptake of calcein acetoxymethyl ester by attached cells. Cells treated with CQ⅐Cu 2ϩ revealed significantly lower adhesion to collagen type IV (*, p Ͻ 0.0001). The loss of adhesion was significantly inhibited (**, p Ͻ 0.001-0.0001) by co-treatment with LY-294,002 ( phase clinical trials (8). CQ levels in the brain may reach 20% of serum levels, which equates to 2. 6 -5 M (20). These concentrations were well within the range of CQ levels found to inhibit A␤ in our cultures if sufficient Cu 2ϩ or Zn 2ϩ was available. Cu 2ϩ levels can range from 1.7 M in the extracellular space of the brain to 250 M in the synaptic cleft, whereas Zn 2ϩ is also highly abundant in the brain, with synaptic levels reaching 300 M (21). Further investigation is required to determine whether CQ can transport other metals (i.e. APP-CHO cells were treated with 10 M CQ⅐Cu 2ϩ for 6 h, and MMP activity was assayed in cell lysates and the conditioned (Cond.) medium. MMP-1 and MMP-8 activities were not significantly altered in cells exposed to CQ⅐Cu 2ϩ . MMP-2 and MMP-3 activities were significantly elevated by CQ⅐Cu 2ϩ (*, p Ͻ 0.001; **, p Ͻ 0.05). No significant effects were observed using a broad-spectrum MMP substrate or a substrate recognized by both MMP-2 and MMP-9. Inset, Western blot analysis of cell lysates using antisera to MMP-2 and MMP-9. Western blotting confirmed that latent (upper band) and active (lower band) forms of MMP-2 were up-regulated in cultures treated with CQ⅐Cu 2ϩ compared with controls. No change in latent MMP-9 (upper band ) was observed in CQ⅐Cu 2ϩ -treated cultures, although a slight increase in active MMP-9 (lower band) was seen. B, effect of inhibitors on MMP activity induced by CQ⅐Cu 2ϩ . APP-CHO cells were exposed to CQ⅐Cu 2ϩ (  Although CQ is neurotoxic in vitro at low concentrations (22), cell lines are relatively more resistant to CQ than are primary neurons, and we found no evidence of increased cell death after 6 h of exposure to CQ and metals. Moreover, AD patients treated with 250 or 750 mg of CQ/day have not revealed complications that would indicate severe neurotoxicity (8). Similarly, mice treated with intraperitoneal injections of 28 mg/kg CQ also failed to show evidence of cytotoxicity (23). These findings suggest that the actions and toxicity of CQ in vivo are likely to be complex and dependent on the availability of "free" metals, antioxidants levels, and cellular resistance.
The mechanism of action by CQ⅐Cu 2ϩ is via activation of the PI3K-Akt pathway and subsequent phosphorylation of JNK and ERK1/2. Although it is common to view PI3K-Akt and JNK/p38 as opposing pathways leading to cell survival and apoptosis, respectively (24), JNK activation can also be potentiated through PI3K activation (25,26), as we have demonstrated here. Activation of both PI3K-Akt and JNK pathways has been reported in AD brain tissue, although the downstream consequences of this activity are not clear (27,28).
PI3K is normally activated in response to cell stresses or growth factors, and metals can activate PI3K in some cell culture models (29). Particularly intriguing was our finding that activation of Akt and JNK by treatment with high Cu 2ϩ levels alone (without CQ) or cisplatin had no effect on secreted A␤ levels, demonstrating that, although up-regulation of these pathways is required, by themselves, they are not able to decrease secreted A␤ levels. It is possible that, after exposure to CQ and metals, elevation of Cu 2ϩ (or Zn 2ϩ ) levels in certain subcellular compartments results in specific modulation of multiple metal-dependent pathways, including PI3K activation (Fig. 9). This is consistent with reports that Zn 2ϩ can activate gene expression by a PI3K-and JNK-dependent process (30). Alternatively, elevated metal levels could promote release of growth factors or other ligands that, in turn, activate PI3K, MAPK, and additional pathways. Interestingly, stimulation of MAPK pathways by metal-mediated growth factor release has been reported in lung epithelial cells (16,31).
A common downstream signaling pathway controlled by PI3K activation involves phosphorylation of Akt and subsequent inhibition of GSK3 through phosphorylation (32). Treatment of cultures with LY-294,002 blocked phosphorylation of both Akt and GSK3␣/␤ by CQ⅐Cu 2ϩ . Inhibition (phosphorylation) of GSK3 can result in abrogation of A␤ production in APP-transfected cells (18), consistent with our findings. However, we found that phosphorylation of GSK3␣/␤ correlated closely with increased JNK phosphorylation. Using inhibitors of GSK3 (LiCl and GSK Inhibitor IX), we demonstrated that increased phosphorylation of GSK3 potentiated JNK activation and subsequent A␤ loss. The mechanism behind this is not clear. As phosphorylation of GSK3 leads to its inactivation, the data suggest that activated GSK3 may inhibit or reduce JNK activation by certain stimuli. Similar effects have been reported previously, where a loss of GSK3 activity potentiated JNK activation by growth factors but not by cell stress (17,33). This is consistent with our data suggesting that MAPK pathways are activated by CQ⅐Cu 2ϩ via non-oxidative mechanisms. Down-regulation of GSK3 activity by CQ⅐Cu 2ϩ could also affect tau phosphorylation, and this should be investigated in appropriate neuronal cell models.
Activation of cell signaling pathways by CQ⅐Cu 2ϩ culminated in upregulation of MMP activity and degradation of extracellular A␤. Fig. 9 shows that the order of events are activation of PI3K-Akt, followed by phosphorylation of GSK3 as well as JNK and ERK. Inhibition of these kinases (Akt, JNK, and ERK) blocked activation of MMPs, so they are upstream of MMP activation. Moreover, inhibitors of the kinases and MMPs blocked the loss of A␤, demonstrating that A␤ loss is downstream of these events. MMP activation is often associated with pathological changes to the cellular microenvironment in the brain, including tumor cell tissue invasion and migration and breakdown of blood-brain barrier permeability during cerebral ischemia. However, MMP activation can also have beneficial functions in the brain, including angiogenesis following ischemia and during axon guidance.
Up-regulation of MMP-2 and MMP-3 by Cu 2ϩ or Zn 2ϩ has been reported previously (16,34), although excess Zn 2ϩ can also inhibit MMP activity (35). In this study, we found that either Cu 2ϩ or Zn 2ϩ complexed to CQ induced activation of JNK, ERK, and p38 and loss of secreted A␤. However, we investigated only CQ⅐Cu 2ϩ complexes in detail, and it remains to be determined whether CQ⅐Zn 2ϩ complexes mediate activation of the same MMPs.
Importantly, MMP-2 and MMP-3 levels can be increased through stimulation of the PI3K-Akt and MAPK (JNK and ERK) pathways (36,37), which is consistent with our findings in CQ⅐Cu 2ϩ -treated cells. Nonspecific Akt and JNK phosphorylation was unable to induce the decrease in secreted A␤ levels, indicating that CQ-delivered Cu 2ϩ has a more complex effect on cell signaling pathways, resulting in up-regulation of MMPs. For example, there are a number of soluble inducers of MMP-2 and MMP-3, including transforming growth factor-␤ and epidermal growth factor, that could be released upon exposure to CQ and metals.
A␤ can be degraded in vitro and in vivo by several proteases, including metalloproteases, neprilysin, insulin-degrading enzyme, and MMPs (4,6,38). These metalloproteases may have important roles in clearance of A␤ in the brain, whereas reduced activity in AD patients could promote amyloid FIGURE 9. Schematic of a proposed mechanism showing how CQ⅐Cu 2؉ mediates reduction of A␤ levels in APP-CHO cell cultures. Solid arrows represent established pathways. Dashed arrows represent proposed pathways. CQ⅐Cu 2ϩ complexes enter the cell by an unknown process. Cu 2ϩ induces PI3K and additional cofactors (?) required for JNK activation. PI3K also activates Akt via phosphorylation, which, in turn, mediates phosphorylation of GSK3. PI3K activates MEK1/2 (not shown), resulting in phosphorylation of ERK1/2. Upon phosphorylation, GSK3 potentiates activation of JNK, which either alone or in concert with GSK3 or other signal factors, up-regulates the activities of MMP-2 and MMP-3. This induces an increase in degradation of extracellular or membrane-associated A␤ by these metalloproteases. Sites of inhibitor action are also shown.
deposition. As thiorphan (neprilysin inhibitor) had no effect on A␤ levels in CQ⅐Cu 2ϩ -treated cultures, increased neprilysin activity is unlikely to be involved in A␤ loss. Although the inhibition of A␤ loss by bacitracin is consistent with insulin-degrading enzyme activity toward A␤, immunoblot analysis of the culture medium revealed no increase in insulin-degrading enzyme levels after CQ⅐Cu 2ϩ treatment (data not shown).
It has been shown previously that A␤-(1-40) and A␤-(1-42) can be degraded by MMP-2 (Lys 16 2Leu 17 , Leu 34 2Met 35 , and Met 35 2Val 36 ) and MMP-3 (Glu 3 2Phe 4 ) (4, 39). MMP-6 is also known to degrade A␤ at Lys 16 2Leu 17 , Ala 30 2Ile 31 , Leu 34 2Met 35 , and Gly 37 2Gly 38 (4,39). Our study supports the MMP-mediated degradation of A␤, as both MMP-2 and MMP-3 were up-regulated in response to CQ⅐Cu 2ϩ treatment, and inhibition of these metalloproteases prevented the loss of secreted A␤. Whether other MMPs (40) and proteases (i.e. aminopeptidases) are also involved in the loss of A␤ induced by CQ⅐Cu 2ϩ is not known. Further investigation will be necessary to fully characterize the A␤ cleavage products in CQ⅐Cu 2ϩ -treated cultures and to identify whether MMP-2-and MMP-3-mediated cleavage is a rate-limiting step in the rapid clearance of secreted A␤.
Whether CQ enhances degradation of A␤ in vivo is not known. MMP expression and distribution are altered in AD brains, and up-regulation of MMP activity occurs in response to A␤ exposure in vitro (41). Although this may result from inflammatory processes, it could also be an attempt to increase A␤ degradation. Several recent studies have shown that increases in central nervous system Cu 2ϩ levels result in lower A␤ levels and reduced plaque deposition (42,43). Moreover, Cherny et al. (10) demonstrated that APP transgenic mice treated with CQ have elevated central nervous system Cu 2ϩ and Zn 2ϩ levels together with reduced A␤ deposition. These reports are consistent with our findings here that elevated Cu 2ϩ or Zn 2ϩ levels can reduce A␤ levels by increasing A␤ degradation. Interestingly, small phase clinical trials of CQ demonstrated lower plasma A␤-(1-42) levels with elevated plasma Zn 2ϩ levels in treated patients (8). This could reflect increased peripheral degradation of A␤ through elevated Zn 2ϩ levels. If so, this would raise the possibility of targeting peripheral A␤ with metal ligands as a means of reducing the total A␤ load.
In summary, our studies indicate a potentially important therapeutic role for induction of MMP activation by metal ligands and subsequent A␤ degradation. If CQ also mediates clearance of A␤ in vivo through activation of A␤-degrading MMPs, these findings will have important implications for the future direction of AD therapeutics based on modulation of metal bioavailability.