Sequestration of Copper from β-Amyloid Promotes Selective Lysis by Cyclen-Hybrid Cleavage Agents*

Decelerated degradation of β-amyloid (Aβ) and its interaction with synaptic copper may be pathogenic in Alzheimer disease. Recently, Co(III)-cyclen tagged to an aromatic recognition motif was shown to degrade Aβ in vitro. Here, we report that apocyclen attached to selective Aβ recognition motifs (KLVFF or curcumin) can capture copper bound to Aβ and use the Cu(II) in place of Co(III) to become proteolytically active. The resultant complexes interfere with Aβ aggregation, degrade Aβ into fragments, preventing H2O2 formation and toxicity in neuronal cell culture. Because Aβ binds Cu in amyloid plaques, apocyclen-tagged targeting molecules may be a promising approach to the selective degradation of Aβ in Alzheimer disease. The principle of copper capture could generalize to other amyloidoses where copper is implicated.

pounds (1.2 eq). The stock solutions were incubated at 37°C for 4 h before use.
Oligomeric A␤ (oA␤) Preparation-Soluble A␤(1-42) oligomers were prepared by dissolving 1.0 mg of A␤ in 400 l of HFIP overnight at room temperature (21). 100 l of the resulting seedless A␤ solution was added to 900 l of Millipore water in an Eppendorf tube. After a 10 -20-min incubation at room temperature, the samples were centrifuged for 15 min at 14,000 ϫ g, and the supernatant fraction (pH 2.8 -3.5) was transferred to a new tube and subjected to a gentle stream of N 2 for 5-10 min to evaporate the HFIP. The samples were then stirred at 500 rpm for 24 -48 h at room temperature.
Tyrosine Intrinsic Fluorescence-Fluorescence spectra were collected using a Hitachi FP-4500 fluorescence spectrophotometer (22,23). An excitation frequency of 280 nm (slit width, 5 nm) was used, and data were collected over 290 -400 nm (slit width, 5 nm). Samples were placed in a four-sided quartz fluorescence cuvette, and data were recorded at room temperature.
Hydrogen Peroxide Assay-The colorimetric H 2 O 2 assay was performed in a 96-well microtiter plate (Amresco), according to an existing protocol (10 ThT Assay-ThT assay was performed by combining 20 l of incubated solution with a 700-l solution of 10 M ThT in 12 mM phosphate buffer (pH 7.4). Fluorescence measurements were recorded in a Hitachi FP-4500 fluorescence spectrometer at room temperature using a 1-cm path length quartz cell. The excitation wavelength was set to 440 nm (slit width, 5 nm), and emission was monitored from 450 to 600 nm (slit width, 5 nm). The percent aggregation is arbitrarily defined at baseline (0%) by the fluorescence of a solution of ThT to which no A␤  was added and at maximum (100%) by the fluorescence after a 2-day incubation of A␤(1-42) alone. A linear relationship between ThT fluorescence and percent aggregation is used for simplicity.
TEM-The TEM samples were prepared by placing 8 l of the incubated solution monitored by ThT assay on 300-mesh formvar-coated copper grids for 2 min before removing excess solution. Then the sample was stained with 1-% fresh tungstophosphoric acid for another 2 min. The grid was blotted on filter paper and allowed to dry before observing the specimen in a JEOL-1200EX electron microscope (JEOL) at 100 kV.
Western Blot Analysis-The samples were boiled with sample buffer for 5 min and electrophoresed on a 10 -20% SDS/ polyacrylamide gel using a Tris/Tricine buffer system. Then the gel was transferred onto membrane, blocked with milk (10% w/v) and probed with WO2 antibody. The membrane was incubated with a secondary polyclonal rabbit anti-mouse IgG conjugated to horseradish peroxidase (Amersham Biosciences) and developed by the ECL detection system (Amersham Biosciences).
MALDI-TOF/MS-Samples were first passed through a reverse phase C18 Ziptip (Eppendorf) to remove salts, 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 Biosystems 4700 Proteomics analyzer MALDI-TOF/TOF operated in reflection mode. Myoglobin was used as an internal standard.
Neurotoxicity Assay-Two kinds of cells were studied for the ability of the hybrid compounds to be tolerated in culture and to rescue A␤ toxicity. The mouse neuroblastoma N2a cells were a gift from H. Xu. The cells were plated in 24-well plates (Corning Costar Corporation) and then incubated for 28 h at 37°C. We added 10 M Cu(II) after 4 h from plating, and the medium was changed to fresh minimal Eagle's medium without serum before commencing treatment. Fresh A␤ (100 M) was incubated for 8 h in minimal Eagle's medium at 37°C before the experiment. Cyc-KLVFF (from 1 to 20 M), and A␤ (5 M) were added into the wells, respectively. MTT was then added to each well, and the plates were incubated at 37°C for 4 h in a humidified CO 2 incubator. Stop solution was added to each well; and after 1 h, dimethyl sulfoxide was added, and the absorbance of the colored formazan product was measured at 490 nm. The rat hippocampal neurons were prepared as described (24). Fresh A␤(1-42) (40 M) was incubated with or without cyclen-hybrid cleavage agents (40 M) for 3 days at 37°C in PBS (pH 7.4) and then diluted 1:4 with culture medium and applied to neurons. The neurons were incubated for 2 days at 37°C and 5% CO 2 , and thereafter the cell viability was evaluated using a MTT assay.
Immunofluorescence Microscopy-Rat hippocampal neurons were treated the same as for the neurotoxicity assays. Fresh A␤(1-42) (100 M) was incubated for 8 h in minimal Eagle's medium at 37°C before the experiment. The final concentration of A␤ was 1 M, and cyc(Cu(II))-KLVFF was present at 10 M. After a 2-day incubation, cells were prepared for microscopy as described (24). (25). We first confirmed that Cu(II) induced a reduction in tyrosine fluorescence of A␤  at 307 nm (Fig. 1B), indicating that there is an interaction between Cu(II) and A␤. When coincubated with cyc-KLVFF, A␤  in the presence of Cu(II) exhibited greater fluorescence (Fig. 1B), indicating that the cyc-KLVFF may have removed the Cu(II) from the A␤. To confirm this capture mechanism, the Cu(II) was incubated with A␤ first, fluorescence was measured, whereupon cyc-KLVFF was added to the sample, and fluorescence was reassayed after 1 h. The addition of cyc-KLVFF completely reversed the quenching of tyrosine fluorescence by Cu(II) (Fig. 1C). We also used the generation of H 2 O 2 by A␤-Cu(II) complexes as an index of copper interaction with A␤(1-42) (26). Cyc-KLVFF (0.5 mol eq) inhibited the amount of H 2 O 2 generated by Cu(II)-A␤(1-42) by Ϸ70% (Fig.  1D). Taken together, the data demonstrate that cyc-KLVFF can competitively capture Cu(II) from Cu(II)-A␤.
MALDI-TOF/MS analysis confirmed that incubation of freshly prepared A␤(1-42) with cyc(Cu(II))-KLVFF for 5-7 days resulted in generation of A␤ fragments (Fig. 3C and supplemental Table S1). The cleavage sites were within 5 residues of the KLVFF motif (Fig. 3D). These data indicate that the appearance of A␤ fragments and the disappearance of bands on Western blot are correlated and establish that cyc(Cu(II))-KLVFF degrades the A␤ peptide.
Cyc-KLVFF Protects Neurons from A␤ Toxicity-We reasoned that cyc(Cu(II))-KLVFF may protect neurons from A␤ toxicity, if it is itself nontoxic. N2a neuronal cells were untreated or pretreated with 10 M Cu(II) for 24 h, and then the culture medium was exchanged, whereupon the cells were incubated for 2 days with medium containing A␤(1-42) alone (5 M) or A␤(1-42) with apocyc-KLVFF. After incubation, neuronal viability was assessed by MTT assay, which revealed that A␤(1-42) was toxic, and toxicity was exaggerated by Cu(II) pretreatment (78% survival at 48 h) (Fig. 4A), in agreement with previous findings (11,26). Apocyc-KLVFF rescued A␤(1-42)-mediated toxicity even at concentrations (1 M) one-fifth those of the A␤  in the culture (5 M ; Fig. 4A). Because apocyc-KLVFF cannot degrade A␤ without Cu(II), these data indicate that apocyc-KLVFF can trap Cu(II) from the cell culture to become active. Additionally, even at cyc-KLVFF concentrations 20-fold greater than the concentration needed to effect maximum rescue, there was no evidence of toxicity (data not shown), indicating that the cyc-KLVFF was tolerated by the primary neurons.
In similar experiments, we also appraised neuronal apoptosis as measured by TUNEL assay, and cell morphology was assessed by tubulin immunohistochemistry. As expected, treatment with A␤(1-42) induced cell apoptosis as well as abnormal cell morphology: neurites became dystrophic, and some neurons detached from the slides (Fig. 4B). Coincubation of A␤(1-42) with cyc(Cu(II))-KLVFF significantly reduced the number of TUNEL-positive cells (Fig. 4C) and prevented changes in cell morphology (Fig. 4B).
To determine whether a nonpeptide drug candidate could act as the A␤ recognition ligand, we examined curcumin, which has a high binding affinity for A␤ (20), conjugated to cyclen. Cyclen-curcumin (cyc-Cur) and cyclen-Gly-curcumin (cyc-GCur) (Fig. 5A) were prepared with Cu(II) bound and found to inhibit A␤(1-42) aggregation with potency similar to cyc-(Cu(II))-KLVFF (Fig. 5C). Cyc(Cu(II))-GCur, had an IC 50 of Ϸ3 M and was slightly more potent than cyc(Cu(II))-KLVFF in this assay. The cyc-Cur conjugates could not be tested for H 2 O 2 inhibition because their color interfered with the assay. MALDI-TOF/MS data again supported the likelihood of A␤ cleavage being the mechanism of action (Fig. 6, D and E, and supplemental Table S1).
Neuronal viability assay indicated that in addition to cyc-(Cu(II))-KLVFF, cyc(Cu(II))-LVFF and cyc(Cu(II))-GCur can significantly protect hippocampal neurons from A␤(1-42) tox-icity (Fig. 6F). The other hybrid compounds showed a trend to rescue. It is not clear why cyc(Cu(II))-KLVFF and cyc(Cu(II))-Cur were not as potent in rescuing the neuronal cells, but there was no precise correlation between neuronal rescue and disaggregation or inhibition of H 2 O 2 (Fig. 5, B and C). The variance in cell culture rescue effects might be caused by different rates of cellular catabolism altering the biological half-lives of the various compounds.

DISCUSSION
The clearance or degradation of A␤ holds promise as a therapeutic approach to AD. The current data explore the potential of small molecules such as cyclen to act as lytic agents in living systems. Such potent hydrolytic activities may be expected to be harmful, but our current data show for the first time that Cu(II)-cyclen conjugates are tolerated in neuronal cell culture for periods of days at concentrations (20 M) that may be expected to be in a therapeutic range. The conjugation of cyclen to an A␤ recognition moiety may have improved the tolerability of the lytic molecule to cells.
Cyclen relies upon binding redox-active metal ions (e.g. Co 3ϩ , Cu 2ϩ ) for its lytic activity. Our data also show that apocyclen conjugates can capture Cu 2ϩ that is bound to A␤ (Fig.  1B), inhibiting aggregation ( Fig. 2A), H 2 O 2 formation (Fig. 1C), and toxicity (Fig. 4A). This is important because ionic cobalt does not exist in biochemical systems (cobalt is not ionized in cobalamin), whereas ionic copper is enriched in several protein aggregates, such as amyloid. Furthermore, bound metal ions may not survive the digestive system or pass across the bloodbrain barrier in oral or parenteral preparations, respectively. The apo-preparation also ensures that the cyclen complex remains inactive until it reaches its peptide target, such as Cubound A␤. Of all biological metal ions, Cu 2ϩ has by far the highest binding affinity to cyclen (K a ϭ 10 Ϫ23 ), more than 20 orders of magnitude greater than its affinity for Mg 2ϩ or Ca 2ϩ , by comparison (29). Therefore, even if the cyclen ring becomes occupied by another metal ion en route to its peptide target (e.g. in the blood), the ability of apocyc-KLVFF to rescue A␤ in the mixed metal ion environment of culture medium (Fig. 4A) suggests that ultimately enough Cu 2ϩ exchanges into the cyclen ring at the peptide docking site to activate the lytic activity.
The data from MALDI-TOF/MS (Fig. 3C) show that the major cleavage sites locate to either sides of the KLVFF, which suggests that the recognition domain of cyc-KLVFF is necessary for the process of cleaving A␤. Cyc-KLVFF was well toler- ated by neurons in culture (Fig. 4A), suggesting that its lytic activity is selective for its target. We hypothesize that in vivo the compound will be inactive until it is brought close to the pathological peptide target by the recognition domain. After docking with the peptide, the cyclen domain is brought close to the Cu 2ϩ binding site of A␤ where it can then commandeer Cu 2ϩ bound to the peptide and become lytically active, so cleaving the peptide mainly within the metal binding domain (30,31), as well as carboxyl-terminal to the KLVFF docking site (Fig. 3, C  and D). Our data also suggest that either monomeric and oligomeric A␤ or both are cleaved (Fig. 3, A and B). The combined effects of these compounds are sufficient to provide protection from A␤ toxicity to cultured neuronal cells (Figs. 4 and 6F).
To our knowledge, this is first report of the use of a synthetic activity to rescue protein aggregate-mediated cellular toxicity. The principle of targeting pathological protein aggregates with a small recognition domain ligated to cyclen could have broad applicability to the many forms of amyloidosis and other diseases associated with protein aggregation, including several neurodegenerative diseases such as Huntington disease. Peripheral amyloidosis, like that formed by ␤ 2 -microglobulin, might be even more readily accessible by the cyclen complexes than brain amyloids, especially as there is no bloodbrain barrier to contend with. Given the relatively slow rate of lysis, high plasma levels may be needed for pharmacotherapy, and therefore this class of drug may be best delivered intravenously. Formal toxicology of an intravenous formulation will determine whether a therapeutic dose of these compounds can be tolerated. To maximize the chances of tolerability, we developed an apocyclen strategy that capitalizes on the peptide's abnormal decoration with ionic copper. Other instances where protein aggregates are induced or decorated by copper include ␤ 2 -microglobulin (47) and transthyretin (48). These might also be accessible targets for apocyclen ligands and warrant further investigation.