A novel tricyclic pyrone compound ameliorates cell death associated with intracellular amyloid-β oligomeric complexes

Abstract

The neurotoxicity of amyloid-β protein (Aβ) is widely regarded as one of the fundamental causes of neurodegeneration in Alzheimer's disease (AD). This toxicity is related to Aβ aggregation into oligomers, protofibrils and fibrils. Recent studies suggest that intracellular Aβ, which causes profound toxicity, could be one of the primary therapeutic targets in AD. So far, no compounds targeting intracellular Aβ have been identified. We have investigated the toxicity induced by intracellular Aβ in a neuroblastoma MC65 line and found that it was closely related to intracellular accumulation of oligomeric complexes of Aβ (Aβ-OCs). We further identified a cell-permeable tricyclic pyrone named CP2 that ameliorates this toxicity and significantly reduces the levels of Aβ-OCs. In aqueous solution, CP2 attenuates Aβ oligomerization and prevents the oligomer-induced death of primary cortical neurons. CP2 analogs represent a new class of promising compounds for the amelioration of Aβ toxicities within both intracellular and extracellular sites.

Abbreviations used

Aβ

amyloid-β protein

Aβ-OC

oligomeric complex of Aβ

AD

Alzheimer's disease

AFM

atomic force microscopy

APP

amyloid-β precursor protein

MTT

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

PBS

phosphate-buffered saline

RU

response units

SPR

surface plasmon resonance spectroscopy

TC

tetracycline

TP

tricyclic pyrone

Studies of amyloid-β (Aβ) toxicity using cell culture models have mainly focused on the consequences of applying synthetic Aβ peptides extracellularly to the culture media and have contributed tremendously to our understanding of Alzheimer's disease (AD; Pike et al. 1993; Lorenzo and Yankner 1994). Recently, a growing body of evidence suggests that the intraneuronal accumulation of Aβ might also be toxic and might be the first step of the amyloid cascade leading to AD (Wirths et al. 2004; Gouras et al. 2005). Accumulation of Aβ in abnormal endosomes is associated with the earliest Aβ elevations in AD (Cataldo et al. 2004). Intracellular Aβ was found to be at least 100 000 times more toxic than extracellular Aβ (Kienlen-Campard et al. 2002; Zhang et al. 2002). The intraneuronal oligomeric form of Aβ has been linked to neuritic damage and synaptic alterations (Takahashi et al. 2004; Almeida et al. 2005). Studies of several transgenic mouse models also demonstrated early intraneuronal Aβ deposition before the first appearance of amyloid plaques or behavioral deficits (Shie et al. 2003; Casas et al. 2004; Billings et al. 2005). Intraneuronal Aβ was correlated with the earliest behavioral deficits in a line of transgenic mice (Billings et al. 2005). Taken together, these data point out possible therapeutic benefits from cell permeable compounds that target intraneuronal Aβ (Walsh et al. 2002).

A major obstacle has been the paucity of cell culture models showing a convincing link between intracellular Aβ and cell death, which would be simple enough to be used as a screening tool for potentially therapeutic compounds. Established models relied on long-term viral expression of the amyloid-β precursor protein (APP) in rat cortical neurons (Kienlen-Campard et al. 2002) and micro-injection of synthetic Aβ peptides or an Aβ-expressing cDNA into human cortical neurons (Zhang et al. 2002). A recent model using primary neurons from Tg2576 APP transgenic mice shows Aβ oligomers accumulating in neuronal processes (Takahashi et al. 2004). However, the preparation of this model is relatively time-consuming and the toxicity requires detection by electron microscopy.

By contrast, the MC65 cell line can be conveniently used to screen bioactive compounds (Sopher et al. 1994; Jin et al. 2002; Maezawa et al. 2004; Woltjer et al. 2005). MC65 is a line of human neuroblastoma that conditionally expresses C99, a 99-residue carboxyl terminal fragment derived from the β-secretase cleavage of APP. C99 is subsequently cleaved by γ-secretase to generate Aβ. Following transgene induction, significant loss of cell viability occurs after 68 h. The death is not caused by factors in the media, including secreted Aβ (Sopher et al. 1994). Compared with all the models mentioned above, MC65 cells are easily propagated and the cell toxicity is measured quantitatively by a simple 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. However, the exact intracellular C99-related mechanism leading to cell death is not known.

Previously, we showed that several tricyclic pyrones (TPs) protected MC65 cells from death (Jin et al. 2002; Hua et al. 2003). However, the mechanism of this protection is not known. TPs were first synthesized based on the structures of pyripyropene A, a potent acyl-CoA : cholesterol O-acyltransferase inhibitor (Omura et al. 1993), and arisugacin, a potent acetylcholinesterase inhibitor (Omura et al. 1995; Hua et al. 1997a). Both activities have been shown to affect Aβ production or aggregation (Inestrosa et al. 1996; Hutter-Paier et al. 2004). However, a few TPs protect MC65 cells at concentrations at least 1000-fold lower than those required for acyl-CoA : cholesterol O-acyltransferase and acetylcholinesterase inhibition (Hua et al. 2003), therefore these two activities cannot explain the MC65 protective effect. Here, we demonstrate that a lead TP named CP2 is cell permeable and is able to block Aβ aggregation, effects that may underlie its MC65 protection.

Materials and methods

Chemicals

CP2 and its analogs were prepared according to our published procedures (Hua et al. 1997a, 2003) and purified using a preparative HPLC system [Jupiter C18, 10-μ column (Phenomenex, Torrance, CA, USA); running solvent system: H₂O:CH₃CN:CF₃CO₂H = 20 : 80 : 0.1]. The TPs were identified by ¹H and ¹³C NMR, IR, MS (matrix-assisted laser desorption ionization and electrospray ionization), and elemental analyses. L-685 458 <{1S-benzyl-4R-[1-(1S-carbamoyl-2-phenethylcarbamoyl)-1S-3-methylbutylcarbamoyl]-2R-hydroxy-5-phenylpentyl}carbamic acid tert-butyl ester> was purchased from Calbiochem (San Diego, CA, USA). Chrysamine G was a gift from Dr Hank Kung. Curcumin was purchased from Sigma (St Louis, MO, USA). Aβ1–40 and Aβ1–42 peptides were purchased from Bachem Bioscience (King of Prussia, PA, USA).

Cell cultures

MC65 cells were grown in the presence of 1 μg/mL tetracycline (TC) as described (Sopher et al. 1994; Jin et al. 2002; Maezawa et al. 2004). The cell toxicity was induced by the removal of TC to induce C99 expression. To do so, the cells were washed extensively, and plated at a density of 1.2 to 1.5 × 10⁵ cells/cm² in Opti-minimal essential medium (MEM; without phenol-red) from Gibco/BRL (Carlsbad, CA, USA) without serum and without TC. The cytotoxicity was determined using a colorimetric MTT assay, the results of which were comparable with data obtained using counts of viable cells based on trypan blue exclusion and the live/dead assay (Maezawa et al. 2004; Woltjer et al. 2005). The primary cultures of cortical neurons from newborn mice were performed as described (Jin et al. 2004). The N2a-APPsw neuroblastoma line expressing APP with the Swedish mutation was a gift from Dr Sangram Sisodia (Thinakaran et al. 1996). The preparation of cell homogenates and western blotting were performed as previously described (Jin et al. 2004; Maezawa et al. 2004).

Preparation of unaggregated and oligomeric Aβ solutions

Solutions of seedless, unaggregated Aβ and oligomeric, non-fibrillar Aβ were prepared according to established protocols (Dahlgren et al. 2002; Chromy et al. 2003; Kayed et al. 2004). Briefly, dried Aβ peptides were first dissolved in ddH₂O with 0.1% trifluoroacetic acid. The solution was dialyzed against 0.1% trifluoroacetic acid and lyophilized. Lyophilized peptides were then completely dissolved in hexafluoroisopropanol at 25°C for 1–3 h and lyophilized. The resulting Aβ film was dissolved with dimethylsulfoxide (DMSO) and stored at −20°C. To generate unaggregated (mostly monomeric) Aβ solutions, the Aβ stock in DMSO was diluted directly into phosphate-buffered saline (PBS). To make oligomers, the 10-μm unaggregated Aβ solution was incubated at 4°C for 48 h with stirring at 300 r.p.m. The resulting oligomers were verified by atomic force microscopy as described (Dahlgren et al. 2002; Chromy et al. 2003; Kayed et al. 2004), aliquoted, and stored at −20°C. As a control, scrambled Aβ peptides did not form oligomers, as expected.

Sandwich ELISA for Aβ

The assay for Aβ monomers in the conditioned media and in the guanidine HCl extracts of cells was performed according to our published protocols (Jin et al. 2004).

Immunofluorescent staining and confocal microscopy

Immunofluorescence labeling and confocal microscopy for Aβ-immunoreactive aggregates were performed according to our published protocols (Jin et al. 2004; Maezawa et al. 2004). The rabbit polyclonal antibody specific for Aβ40 was purchased from Calbiochem.

Cell permeation of [¹⁴C]CP2

[¹⁴C]CP2 was prepared from ¹⁴C-labeled adenine (Amersham Pharmacia Biotech, Piscataway, NJ, USA; Hua et al. 1997a, 2003), with a specific activity of 0.38 μCi/mg. [¹⁴C]CP2 (0.04 μCi) was added to culture wells of 2.5 × 10⁶ cells and the cultures incubated for the indicated periods of time. MTT and live/dead assays of parallel cultured cells thus incubated showed no observable toxicity. The cells were scraped and spun down. The cell pellets were washed twice with ice-cold PBS and then trypsinized for 10 min on ice. Afterwards, the supernatants were kept as ‘trypsinizable cell surface fraction’ and the washed pellets as ‘non-trypsinizable intracellular fraction’. Both fractions were subjected to scintillation counting.

Surface plasmon resonance spectroscopy (SPR)

A Biacore 3000 (Biacore Inc., Piscataway, NJ, USA), equipped with four flow cells in one Sensor chip, was used for real-time binding studies. The Aβ1–40 and Aβ1–42 peptides were separately immobilized on gold sensor CM5 chips as described (Tjernberg et al. 1996). First, carboxymethylated dextran of CM5 was activated with N-hydroxysuccinimide in pH 7.4 buffer, followed by the coupling with Aβ peptides. The resulting Aβ-bound Sensor chip was used in the SPR studies, and the amounts of bound Aβ were measured from the SPR graph. The reference cells were prepared following the same procedure but without addition of Aβ peptides. TPs were dissolved in 50 mm Tris buffer to a concentration of 800 μm and injected into flow cells. The chip surface was exposed to compounds for 60 s, then to running buffer (10 mm HEPES, 150 mm NaCl, 3.4 mm EDTA, 0.05% surfactant P20, pH 7.4) for 180 s. The response unit (RU) was recorded, converted to pmol bound compound, and analyzed. The flow cells were then regenerated with regenerating buffer containing urea, while leaving the immobilized Aβ peptides intact. The same chips with the same amounts of bound Aβ peptides were used for consecutive analysis of different compounds and comparisons were made only between the bindings to the same chips. The amounts of bound compounds were calculated by the ΔRU value, obtained by subtracting the RU at the time before compound injection from the RU at the time just after the injection of running buffer. The ΔRU obtained from the reference cell was considered the background binding and was subtracted before analysis. The kinetic analysis of sensograms from interaction of compounds with the immobilized Aβ was performed using Biaevaluation software, version 3.0 (Biacore Inc., Piscataway, NJ, USA), by fitting the binding curves to a 1 : 1 Langmuir binding model. The dissociation constant K_D (M) is derived from the equation, K_D = k_d/k_a, where k_d and k_a are dissociation- and association-rate constants, respectively.

Atomic force microscopy (AFM)

A Nanoscope IIIa SPM AFM (Digital Instruments Inc., Santa Barbara, CA, USA) was used. Exactly 10-μL samples were aliquoted from reaction mixtures and immediately spotted on freshly cleaved micas. The micas were washed with water twice, and dried with argon. AFM images were collected using a tapping mode (http://www.veeco.com/) with a high-aspect ration tip (Veeco Nanoprobe TM tips, Model TESP-HAR; Nanoscience Instruments, Inc., Phoenix, AZ, USA). The J-imaging program at an 8-bit resolution was used to carry out grain analysis to collect quantitative data. The image was adjusted by setting a threshold so that no noise on the image was present. The same threshold was used for all other images in the same independent experiment. For each experimental condition, 25-μm² images from three independent AFM scans were used to determine the density (particles/field) of oligomers.

Statistics

We examined the statistical significance of differences between groups by applying one-way anova with post-hoc Tukey test or Bonferroni tests, using the SigmaStat 3.1 (Systat Inc. Point Richmond, CA, USA) software program.

Results

MC65 cell death is related to the intracellular generation and aggregation of Aβ

To avoid detection of changes secondary to cell toxicity, we obtained MC65 cell homogenates from early time points after TC removal when the cells were morphologically normal. Western blot analyses demonstrated that C99 expression was followed by other Aβ-containing fragments, of which the 8, 16.5 and 25 kDa proteins predominated (Fig. 1a). Intriguingly, no intracellular monomeric Aβ was detected by Tricine gel analysis. However, our previous studies using Tris/bicine/urea gel electrophoresis of cell homogenates and sandwich ELISA on guanidine-extracted cellular Aβ clearly detected Aβ monomers (Jin et al. 2002, 2004), suggesting the possibility that intracellular Aβ monomers were incorporated into aggregates which were dissociated by urea or guanidine.

Figure 1.  MC65 cell death is dependent upon Aβ generation and closely related to Aβ aggregation into Aβ-OCs. (a) The C99 was induced by removal of TC. At indicated times, the cells were homogenized. Cellular proteins (5 μg protein each) were subjected to Tris/tricine sodium dodecyl sulfate – polyacrylamide gel electrophoresis and western blot analysis with antibody 6E10 (for Aβ1–17). Aβ Std, synthetic Aβ1–40 and Aβ1–42, serving as standard Aβ monomer peptides. (b, c) Aβ generation was inhibited by a γ-secretase inhibitor L-685 458 (2 μm), but not by CP2 (10 μm) or TP17 (10 μm). Cells were treated with the indicated compounds for 48 h. The Aβ levels in the conditioned media (CM) were measured by sandwich ELISA. Error bars represent standard error. (b) MC65 cells. –TC, without TC to induce C99 expression; +TC: with TC to suppress C99 expression. n = 3, **p < 0.001 compared with the –TC group. (c) N2a-APPsw cells. n = 4, ***p < 0.001, compared with the no-treatment control. (d) L-685 458 of indicated concentrations was added at the same time as TC removal. At 72 h, viability was assessed by MTT assay. Data are expressed as mean percentage viability (n = 7) with parallel +TC cultures set at 100% viability. Error bars represent standard error. ***p < 0.001, **p < 0.01, compared with untreated –TC controls. (e) L-685 458 of indicated concentrations was added at the same time as TC removal. Cellular proteins from 24-h cultures were analyzed by western blot using 6E10 (left panel). The quantification of the expression of p8, 16.5 kDa Aβ-OC and 25 kDa Aβ-OC was determined after being normalized to the actin control by measuring the optical density of respective bands using NIH Image. The data are presented as percentage of change ± SEM (n = 3) after treatment with L-685 458 of the indicated concentrations (right panel). (f) The same amounts of proteins from –TC MC65 cell homogenates before (–) and after (+) formic acid (FA) extraction were analyzed by western blot using 6E10. (g) The same amounts of protein from +TC, –TC, and –TC MC65 cells homogenized in the presence of the indicated concentrations of Congo red (CR) were analyzed by western blot using 6E10. (h) The effect of L-685 458 (2 μm) was compared with CP2 (1 μm). Proteins from cells with indicated treatments were analyzed by western blots. Parallel blots were analyzed with 6E10 and B994, respectively, for comparison. A blot probed with β-actin antibody was used to demonstrate equal protein loading.

To test this possibility, we treated MC65 cells with a cell-permeable, highly specific γ-secretase inhibitor, L-685 458 (Doerfler et al. 2001). This treatment efficiently blocked Aβ production from both MC65 cells (Fig. 1b) and N2a-APPsw cells (Fig. 1c). L-685 458 also protected MC65 cells in a dose-dependent manner (EC₅₀ = 110 nm; Fig. 1d), suggesting that generation of Aβ was required for toxicity. At the protective concentrations, L-685 458 reduced the 16.5- and 25-kDa bands in a dose-dependent manner (Fig. 1e), suggesting that these bands represent homo- or hetero-oligomeric complexes of Aβ (Aβ-OCs). Extraction of cell homogenates with formic acid completely disaggregated Aβ-OCs and yielded monomeric Aβ (Fig. 1f). To exclude the possibility that Aβ-OCs were formed during the homogenization process, MC65 cells were homogenized in the presence of Congo red to inhibit further aggregation of Aβ (Lorenzo and Yankner 1994; Podlisny et al. 1998). This modification did not reduce the expression of Aβ-OCs (Fig. 1g). Collectively, these data suggest that, in MC65 cells, the monomeric Aβ either rapidly aggregated into Aβ-OCs or were secreted.

Previously, it was concluded that the 8-kDa band (p8) consists of Aβ dimers (Woltjer et al. 2005). However, the observation that it was only partially reduced by L-685 458 treatment (Fig. 1e) and by formic acid disaggregation (Fig. 1f) indicates that it was not entirely composed of Aβ peptides. As p8 was not recognized by antibody B994 for the carboxyl terminal 39 amino acids of APP (Jin et al. 2004) (Figs 1h and 3a), it lacks the most carboxyl terminal APP residues. p8 is therefore likely to be unresolved bands of Aβ dimers and CTFΔ31, a caspase cleavage product of C99 that contains the Aβ domain (Lu et al. 2000). These interpretations require clarification by further studies.

Figure 3.  CP2 inhibits the formation of Aβ-OCs. (a) MC65 cells were treated with the indicated compounds for 24 h (two micromolar N9′-analog and TP17 were used). Western blots of the same set of cellular proteins were analyzed with 6E10, B994, and anti-β actin. (b) MC65 cells were treated with TC or without TC, but with the indicated concentrations of CP2 for 24 h. Cellular proteins were analyzed by western blots using 6E10. (c) Twenty-four hours after TC removal and treatment with the indicated compounds (2 μm), MC65 cells were fixed, permeabilized, immunolabeled with anti-Aβ40 (green fluorescence) and counterstained, followed by confocal microscopy. Arrows point to Aβ40-immunoreactive coarse deposits.

To explore the possibility that Aβ-OCs and p8 might be attaching themselves to the cell surface rather than intracellularly, we subjected the intact cells to brief trypsinization to remove Aβ adherent to the cell surface prior to cell homogenization (Walsh et al. 2000). This procedure did not alter the detection of Aβ-OCs and p8 (data not shown). In addition, confocal images of MC65 cells after TC removal did not show a cell membrane profile of Aβ-immunoreactive deposits (Fig. 3c). These findings, taken together with a previous observation that extracellular Aβ cannot be induced to form oligomers, strongly suggest that the observed Aβ-OCs and p8 are accumulated intracellularly.

Western blots probed with antibody B994 revealed that MC65 cells also expressed C83, an α-secretase cleavage product of C99 that is not recognized by 6E10 (Fig. 1h, right panel). Interestingly, L-685 458 treatment substantially increased the level of C83 (a 419 ± 109% increase with 2 μm L-685,458, n = 3), suggesting that inhibition of the γ-secretase might shunt the degradation of C99 toward the α-secretase pathway. None of the Aβ-OCs were immunoreactive to B994 (Figs 1h and 3a), indicating that the carboxyl terminal 39 amino acids of APP were not involved in Aβ-OC formation.

CP2 inhibits the formation of Aβ-OCs and protects MC65 cells

To further analyze the mechanism of MC65 death, we treated MC65 cells with several analogs of TPs. We had identified CP2, a candidate TP with a strong MC65 protective effect and low toxicity (Jin et al. 2002; Hua et al. 2003; Fig. 2). Interestingly, the MC65 assay was sensitive enough to differentiate minor alterations of the CP2 structure. For example, while both carbon-12-R- and 12-S-isomers of CP2 showed indistinguishable strong effects (EC₅₀ = 0.15 μm), the N9′-analog regioisomer of CP2 was less active (EC₅₀ = 3.0 μm) than CP2. The N10′-analog regioisomer and the flat tricyclic core structure TP17 were entirely inactive (Fig. 2). CP1 was referred to as compound 4 in Hua et al. (1997a), and fully protected MC65 cells but was less effective (EC₅₀ = 2.6 μm) than CP2. These results imply the existence of a cellular substrate that interacts with CP2 to engender a strict structure–activity relationship.

Figure 2.  Chemical structures and MC65 protection activities of selected TPs. (a) Chemical structures, not drawn to scale. The two diasteromers of CP2 at carbon-12 are indicated by C12R-CP2 and C12S-CP2. The two regioisomers of CP2 are referred to as N9′-analog and N10′-analog. (b) MC65 protection assay was performed as described in Fig. 1(d). Shown are the means from three to 10 experiments. Error bars are not shown. (c) Phase-contrast photomicrographs of MC65 cells 3 days after incubation in the indicated conditions.

We asked whether Aβ might be one of the targets of CP2. We found that CP2, but not the N9′-analog at equivalent concentrations or TP17 at all concentrations, reduced the levels of cellular Aβ-OCs in a dose-dependent manner (Figs 3a and b), similar to L-685 458 (Fig. 1g). None of the TPs consistently influenced the levels of C99 and C83 (Fig. 3a). When treated with 2 μm CP2, there were > 95% reductions of the 16.5- and 25-kDa bands, but only a 32.1 ± 6.5% (n = 4) reduction of p8, similar to the effect of L-685 458 and formic acid disaggregation. This result suggests that CP2 treatment also resulted in disaggregation of Aβ dimers. The parallel effects of CP2 and L-685 458 on MC65 cell viability and Aβ-OC levels suggest that the aggregation of Aβ into Aβ-OCs might play a major role in events leading to cytotoxicity in MC65 cells. Unlike L-685,458, however, CP2 did not decrease Aβ levels (Figs 1b and c), suggesting that the effect of CP2 might be because of a direct modification of Aβ assembly.

Previously, we showed that MC65 cell death was preceded by the appearance of cytoplasmic Aβ-immunoreactive coarse granular deposits, using 4G8 antibody (Maezawa et al. 2004). As 4G8 also recognizes APP and C99, we extended this characterization by using an antibody specific to Aβ40, but not recognizing APP or C99. As expected, +TC cells had no visible Aβ40 immunoreactivity, while 24 h after TC removal the cells showed abundant cytoplasmic Aβ40-immunoreactive aggregates (Fig. 3c). We found that CP2, but not TP17, reduced the level of cytoplasmic Aβ40-immunoreactive deposits. This result confirms that the ability of CP2 to disrupt or prevent the formation of intracellular Aβ aggregates is closely related to its MC65 protective effect.

CP2 penetrates the cells and interacts with Aβ

Our working hypothesis is that CP2 penetrates the cells, binds to Aβ or Aβ-OCs, and modifies their assembly. Applying the rule of Lipinski et al. (2001), CP2, having one hydrogen-bond donor, seven hydrogen-bond acceptors, a molecular weight of 393, a measured octanol/water partition coefficient (P_Oct) of 159 and a calculated Log P_Oct value of 2.2, is predicted to have good permeation properties. Indeed, by incubating MC65 cells with radiotracer [¹⁴C]CP2, we found a rapid accumulation of radioactivity inside the cells (Fig. 4). At 2 h after the addition of [¹⁴C]CP2, the levels of cell-associated CP2 (both cell surface, trypsinizable CP2 and intracellular, non-trypsinized CP2) were saturated. In particular, intracellular CP2 level was maintained at around 450 pmol/10⁵ cells, which was equivalent to 2.7 × 10⁹ molecules per cell.

Figure 4.  CP2 rapidly accumulates inside the cells. MC65 cells were incubated with [¹⁴C]CP2 for the indicated periods of time, washed, and the cell-associated counts that were trypsinizable from cell pellets (cell surface CP2, empty circles) and those remaining in cells after trypsinization (intracellular CP2, filled circles) were quantified. Expressed values are quantities of CP2 calculated from the radioactivity of 10⁵ cells.

To test whether CP2 directly interacts with Aβ, we performed direct binding studies using Biacore's surface plasmon resonance spectroscopy (SPR; Tjernberg et al. 1996; Karlsson 2004; Myszka 2004). We noticed that Aβ42 invariably aggregated to a certain extent during the preparation of chips, therefore the observed binding of compounds to chips immobilized with Aβ42 may not fully represent the binding of compounds to monomeric Aβ42 and was always smaller than the binding to Aβ40, which rarely aggregated during preparation. Despite this limitation, the binding of CP2 to Aβ40- and Aβ42-coated chips was clearly demonstrated, although the responses appeared small because of the small size of CP2 (Fig. 5). On average (n = 3), approximately 1.5 equivalents of CP2, 0.46 equivalents of the N9′-analog, but no TP17 were found to bind one equivalent of Aβ1 40, respectively. Similarly, approximately 1 equivalent of CP2, 0.2 equivalents of the N9′-analog, but no TP17 were found to bind one equivalent of Aβ1 42, respectively. The results of kinetic analysis of the sensograms resulting from the interactions of compounds with immoblilized Aβ in a physiological buffer are shown in Table 1. CP2 exhibited a very high binding affinity to Aβ40 (K_D = 5.05 nm), and a high affinity to Aβ42 (K_D = 269 nm). Although the N9′-analog showed an intermediate amount of binding to the chips, it exhibited K_D values equivalent or larger than those of CP2. TP17 showed no binding.

Figure 5.  Real-time detection of TP binding to Aβ1–40 (a) and Aβ1–42 (b) by SPR. SPR was performed as described in Materials and methods. The results from a typical experiment are shown. The response in the reference cell (without immobilized Aβ) was traced by the blue line and that in the experimental cell (with immobilized Aβ) by the red line. Note that the y-coordinates showing the SPR response units are on different scales between (a) and (b). See text for calculated binding. N9TP, N9′-analog.

Table 1.   The values of dissociation constant K_D for binding of compounds to Aβ peptides immobilized on SPR chips

Analyte	Aβ40	Aβ42
The KD (M) is derived from the equation, KD = kd/ka where kd and ka are dissociation- and association-rate constants, respectively. NB, no binding.
CP2	5.05 × 10^−9	2.69 × 10^−7
N9 analog	9.96 × 10^−9	2.44 × 10^−8
TP17	NB	NB

CP2 attenuates Aβ oligomerization

Through direct interactions with Aβ peptides, CP2 may inhibit the formation of Aβ oligomers, which might be the primary toxic Aβ species (Podlisny et al. 1998; Dahlgren et al. 2002; Walsh et al. 2002; Chromy et al. 2003; Kayed et al. 2004). To test this possibility, we followed the aggregation states of Aβ by scanning with an AFM. When starting from seedless, unaggregated Aβ solutions, Aβ40 peptides formed a few spherical oligomers at 24 h and protofibrils at 72 h. The Aβ42 peptides showed faster aggregation, with numerous spherical oligomers and occasional protofibrils at 24 h, and were mostly precipitated out at 48 h (Fig. 6a). Section analysis of AFM images showed that most Aβ oligomers at 24 h measured 5–7 nm in z-height, suggesting molecular weights of ∼100 kDa (Dahlgren et al. 2002). When CP2 was added to the starting Aβ solution, it attenuated the oligomerization of both Aβ40 and Aβ42 at a 1 : 1 molar ratio (Figs 6b and c), suggesting that it bound to a site or sites critical for Aβ aggregation. This effect was not seen by TP17.

Figure 6.  CP2 attenuates Aβ aggregation in aqueous solution. (a) Seedless, unaggregated Aβ peptides show time-dependent aggregation in PBS. Representative 1-μm² AFM images of 10-μm Aβ1–40 and Aβ1–42 aggregation reactions at the indicated times are shown. Scale bar, 0.25 μm. (b) The addition of 1 equivalent of CP2, but not TP17, to the starting Aβ solutions (10 μm) attenuated formation of Aβ oligomers. Representative 4-μm² AFM images from 24-h solutions are shown. Scale bar, 0.5 μm. (c) Oligomer densities in AFM images of samples from the 24-h incubation were quantified using the Image J Program. n = 3, *p < 0.01, **p < 0.001, compared with their respective no-compound controls (no cpd).

CP2 blocks the toxicity induced by extracellularly applied Aβ oligomers

We asked whether CP2 treatment might also affect the toxicity induced by extracellularly applied Aβ oligomers. To test this possibility, we made Aβ42 oligomers according to established protocols (Dahlgren et al. 2002; Chromy et al. 2003; Kayed et al. 2004) and applied them to primary cortical neurons. Comparable with previous data using N2a neuroblastoma cells (Dahlgren et al. 2002), Aβ42 oligomers induced a small degree of toxicity at 10 nm, and statistically significant toxicity starting at 50 nm(Fig. 7a). In contrast, unaggregated Aβ42 induced no significant toxicity at tested concentrations. When co-administered with 50 nm Aβ42 oligomers at a 1 : 1 molar ratio, CP2, but not TP17, completely blocked neuronal toxicity (Fig. 7b). This result demonstrates that, CP2, although initially selected based on its protection from cell toxicity related to intracellular Aβ, is able to block the toxicity induced by extracellularly applied Aβ oligomers.

Figure 7.  CP2 protects primary cortical neurons from Aβ42 oligomer-induced toxicity. (a) Primary mouse cortical neurons (8 days in culture) were treated with indicated concentrations of Aβ42 monomer or oligomer preparations. After 48 h, neuronal viability was assessed by the MTT assay. Data are expressed as mean percentage viability with no treatment controls set at 100% viability. Error bars represent standard error. n = 3, *p < 0.001 compared with the no-treatment control. (b) Neurons were treated with 50 nm monomers (M), oligomers (O), or oligomers in the presence of 50 nm of CP2 (O + CP2) or TP17 (O + TP17). Neuronal viability was assessed as in (a). n = 4, *p < 0.001 compared with the no treatment control, **p < 0.001 for comparisons between O and O + CP2, and between O + CP2 and O + TP17. p > 0.05 when O and O + TP17 were compared.

Discussion

Intracellular aggregation of proteins is increasingly being linked to neurodegenerative processes and occurs in increasing numbers of diseases collectively called conformational diseases. The mechanism by which intracellular Aβ misfolds, aggregates, and becomes toxic to neurons remains poorly understood. Cell models and inhibitor molecules would contribute to our understanding of this mechanism. We demonstrated that MC65 cells could serve as a model for studying the death of neuron-like cells as a result of intracellular Aβ. MC65 cell death was closely associated with and likely dependent upon the intracellular aggregated forms of Aβ. We took advantage of these properties and successfully selected CP2 that was cell permeable and capable of inhibiting intracellular Aβ aggregation into Aβ-OCs and extracellular Aβ aggregation into oligomers. Our results validate the usefulness of the MC65 protection assay for the screening of small molecules with anti-Aβ aggregation properties.

The rescuing effect demonstrated by the γ-secretase inhibitor clearly indicates that MC65 death is dependent on the production of Aβ. Based on our ELISA quantification of secreted Aβ and Aβ in the guanidine extracts of MC65 cells (Jin et al. 2004), the intracellular Aβ, on average, was only approximately 6% of the total Aβ produced. However, probably because of its high focal concentrations, its strong membrane toxicity (Kayed et al. 2004), or the contributions from unknown cellular co-factors, intracellular Aβ is toxic to cells despite its small quantity. Furthermore, the observations that both L-685 458 and CP2 reduced Aβ-OCs, but CP2 did not reduce the overall Aβ level, strongly suggest that Aβ-OCs might be the major cause of MC65 cell death. Although other indirect effects on intracellular Aβ aggregation/toxicity are possible, our data strongly suggest that a direct effect of CP2 on intracellular Aβ aggregation may, at least in part, underlie CP2′s MC65 protective effect.

Previously, we showed that antioxidants fully protect MC65 cells (Sopher et al. 1996; Maezawa et al. 2004). We excluded the possibility that CP2 was an antioxidant because: (i) the cyclic voltammogram revealed that the oxidation potential E_p,a of CP2 was +1.1 V, too high to be considered an antioxidant (de Souza et al. 2003). By comparison, under similar experimental conditions, the E_p,a of ferrocene was +0.5 V. Typical antioxidants, such as flavonoids, have E_p,a values between +0.58 and +0.91 V; (ii) the HT-22 oxidative stress assay (Tan et al. 1998) showed no protective effect by CP2; (iii) CP2 did not protect from cell toxicities as a result of H₂O₂ treatment or culturing in 95% oxygen. Furthermore, CP2 was not effective in rescuing cell death in apoptotic death models such as serum withdrawal from PC12 cells or nerve growth factor withdrawal from differentiated PC12 cells (data not shown) (Greene 1978). Therefore, it appears that CP2 is not a general antioxidant nor an anti-apoptotic agent.

To our knowledge, CP2 is the first characterized small molecule that inhibits intracellular Aβ aggregation and blocks intracellular Aβ toxicity. Several small molecules, such as azodye derivatives [for example, chrysamine G (CG) and RS-0406], curcumin and DAPH (4,5-dianilinophthalimide), have been characterized based on their ability to inhibit the aggregation of synthetic Aβ peptides in aqueous solution (Blanchard et al. 2004; Walsh et al. 2005; Yang et al. 2005). It remains to be shown whether they block intracellular Aβ aggregation, a process intimately related to membrane/vesicle structures. The azodyes and curcumin share a symmetrical flat structure that allows alignment with the β-pleated sheet (Yang et al. 2005), thus providing their anti-aggregation effect. By comparison, CP2 has an asymmetric structure, while TP17, a flat structure as revealed by X-ray analysis (Hua et al. 1997b), is entirely inactive in our assays. Therefore, CP2 appears to have an anti-aggregation property distinct from that of azodyes and curcumin. Interestingly, we found that CG efficiently attenuated Aβ aggregation in an aqueous solution, similar to CP2. However, neither CG nor curcumin offered MC65 protection (data not shown), despite their previously suggested antioxidation and anti-Aβ effects (Lim et al. 2001; Ishii et al. 2002). Therefore, our data raise an interesting possibility that the mechanisms of aggregation involving vesicle/membrane-associated, cell-produced Aβ would be distinct from those observed with synthetic Aβ in an aqueous solution. It is highly desirable that, in addition to inhibiting the Aβ aggregation in the aqueous solution, an anti-Aβ compound also inhibits the aggregation of membrane/vesicle-associated Aβ and counteracts the deleterious effects of Aβ on the membrane. Our data suggest that CP2 might possess such desirable properties.

Despite their simple core structure, the tricyclic pyrones promise a boundless potential for a variety of novel compounds, which may be fine-tuned to produce leads with desired properties. As demonstrated in our data, small modifications of CP2 can significantly change the biological and biophysical effects of the compounds. Possible modifications of CP2 include functionalization of the methyl group of the pyrone ring, the C7 alkyl side chain, the C9 allylic methylene, and the C10 double bond, to name a few. Knowledge of the structure–activity relationship based on future CP2 derivatives should help elucidate the poorly understood mechanism of intraneuronal Aβ aggregation and toxicity.

Acknowledgements

This work was supported by grants from the NIH (CA86842 and AG19711), the National Science Foundation (CHE-0078921), the American Chemical Society PRF (40345-AC1), the American Heart Association, Heartland Affiliate (0350652Z), Terry C. Johnson Center for Basic Cancer Research (Kansas State University), and the UC Davis Health Science Research Fund.