|Home | About | Journals | Submit | Contact Us | Français|
Alzheimer’s disease (AD) is characterized by depositions of β-amyloid (Aβ) aggregates as amyloid in the brain. To facilitate diagnosis of AD by radioligand imaging, several highly specific small-molecule amyloid ligands have been developed. Because amyloid ligands display excellent pharmacokinetics properties and brain bioavailability, and because we have previously shown that some amyloid ligands bind the highly neurotoxic Aβ oligomers (AβO) with high affinities, they may also be valuable candidates for anti-Aβ therapies. Here we identified two fluorene compounds from libraries of amyloid ligands, initially based on their ability to block cell death secondary to intracellular AβO. We found that the lead fluorenes were able to reduce the amyloid burden including the levels of AβO in cultured neurons and in 5xFAD mice. To explain these in vitro and in vivo effects, we found that the lead fluorenes bind and destabilize AβO as shown by electron paramagnetic resonance spectroscopy studies, and block the harmful AβO-synapse interaction. These fluorenes and future derivatives, therefore, have a potential use in AD therapy and research.
AD is characterized by deposition of various Aβ aggregates forming amyloid in the brain. To facilitate diagnosis of AD, compounds with various chemical core structures have been developed for their use as small-molecule probes (collectively called amyloid ligands here) for in vivo detection of Aβ plaques in patients with AD (Cai et al., 2007). Candidate amyloid ligands can penetrate the intact blood–brain barrier (BBB) and bind to various cerebral Aβ aggregates with high affinity and selectivity (Cai et al., 2007). However, the functional consequence of binding of these compounds to Aβ aggregates remains unknown.
Aβ aggregates induce various degrees of neurotoxicity and have been hypothesized to be causally related to dementia in AD. A major impediment to the development of effective anti-Aβ compounds for AD therapy is that essentially 100% of large-molecule drugs and >98% of small-molecule drugs fail to cross the BBB (Pardridge, 2007). Because several amyloid ligands display excellent pharmacokinetics properties and brain bioavailability, we consider that they are valuable candidates for the development of AD research and therapeutic agents specifically targeting Aβ aggregates (Lee, 2002; Maezawa et al., 2008). This approach was first proposed by Lee (2002), who showed that some amyloid ligands arrest amyloid fibril formation in vitro. Extending this line of study, we intend to select amyloid ligands that also target AβO at both intraneuronal and extracellular sites and block AβO-induced toxicity. This is because accumulating evidence indicates that AβO, more than Aβ fibrils, can induce severe neurodegeneration and cognitive deficits (Walsh et al., 2002; Kayed et al., 2004; Klein et al., 2004; Lesne et al., 2006; Maezawa et al., 2006; LaFerla et al., 2007). Importantly, we previously showed that some amyloid ligands bind AβO with high affinities, suggesting their potential use in anti-AβO therapy (Maezawa et al., 2008). In addition, data obtained from using human tissue and animal models support that Aβ oligomerization initiates within neurons, indicating the importance of compounds targeting intraneuronal AβO (Walsh et al., 2002; LaFerla et al., 2007). Here we identified two lead fluorene compounds code-named K01-162 and K01-186 based on their ability to block cell death secondary to intracellular AβO. These compounds bind and destabilize AβO, and are capable of penetrating the brain and reducing the cerebral amyloid burden in APP transgenic mice, therefore have a potential use in AD therapy and research.
The line Tg6799 5xFAD mice co-express human APP695 with the Swedish (K670N, M671L), Florida (I716V), and London (V717I) mutations and human PS1 harboring M146L and L286V mutations (Oakley et al., 2006). For intracerebroventricular infusion, miniosmotic pumps (Alzet) were loaded either with 100 µM K01-162, K01-182 or equivalent amount of DMSO solvent (mock) in 100 µl of artificial cerebrospinal fluid. The procedure for implantation of the pumps was the same as previously described (Dolev and Michaelson, 2004). Mice were infused at a flow rate of 0.25 µl/h for 2 weeks. This amounted to ~400 ng/g of brain/day. At the conclusion of the infusion, the mice were sacrifice and their brains were cut in half sagitally. The left hemispheres were snap frozen for biochemical assays. The right hemispheres were fixed in 4% paraformaldehyde for immunohistochemical studies. All animal procedures were approved by the UC Davis Animal Care and Use Committee. All surgical procedures were performed with care to minimize pain and discomfort.
Cultured cells or 1-mm thick cryosections of frozen mouse brains were lyses in radioimmune precipitation assay (RIPA) buffer with protease inhibitor mixture solution, sonicated briefly, and incubated for 20 min at 4 °C. The RIPA-soluble fractions were obtained from the supernatants after centrifuge at 100,000 × g. Western blots and immunofluorescent staining were performed according to our published protocols (Hong et al., 2007; Jin et al., 2004; Maezawa et al., 2006). For each mouse, a 20-µm thick coronal section containing hippocampus and subiculum at 2mm rostral to the posterior end of the cerebrum and a second section at 100 µm further rostral were immunostained and photographed for quantification of amyloid load using the Photoshop and Image J softwares.
The analysis of the compound binding to the AβO (made of Aβ1–42) on SPR sensor chip was performed following our previously published protocol (Maezawa et al., 2008). For details, please see Supplementary data.
Site-specific incorporation of a nitroxide spin label at position 2 of Aβ was achieved by synthesis of Aβ1–40 containing a cysteine substitution at residue 2. For details, please see Supplementary Fig. 1 and Supplementary Methods.
The spin-labeled and unlabeled Aβ1–40 peptides were completely dissolved in hexafluoroisopropanol (HFIP) at room temperature for 1–3 h and lyophilized. The resulting Aβ film was dissolved with DMSO and stored at −80 °C. To generate spin-labeled AβO, a mixture of 10 µM of labeled and 30 µM of unlabeled Aβ1–40 in PBS was incubated for 6 h at 4 °C. The resulting AβO were confirmed by AFM and their cytotoxicity to N2A neuroblastoma cells (Hong et al., 2007) was found comparable to the unlabeled AβO.
EPR measurements were carried out in a JEOL X-band spectrometer fitted with a loop-gap resonator. To measure EPR spectra, 9 µl of spin-labeled oligomer solution was combined with 1 µl of 100 µM compound in DMSO (or 1 µl of DMSO vehicle control), incubated for 1 h at 4 °C, and then an aliquot (~5 µl) transferred to a sealed quartz capillary and then placed in the resonator. Spectra were acquired at room temperature (20–22 °C) by signal-averaging three 60-s scan over a field of 100G at a microwave power of 2 mW and a modulation amplitude optimized to the natural line width of the individual spectrum (typically 0.5 G).
AFM and EM were performed as previously described (Hong et al., 2007).
The assay for the antioxidative effect of compounds was performed according to published protocols (Mishra et al., 2006) with modification. Briefly, in a 100 µl of solution, HRP (100 ng) was mixed with hydrogen peroxide (0.03%) in the absence or presence of different concentrations of compounds (16–1000 nM). The reaction was initiated by adding 50 µl of 20 µM o-phenylenediamine dihydrochloride (OPD) substrate. After 5 min incubation at room temperature, the reaction was stopped by adding H2SO4 and absorbance at 490 nm was measured.
We examined the statistical significance of differences between groups by applying one-way analysis of variance (ANOVA) with post hoc Tukey test, using the SigmaStat 3.1 (Systat Inc. Point Richmond, CA) program.
Because the primary goal of our compound development is neuron protection, we employed our previously developed neuronal culture-based assay called MC65 protection assay for primary screening (Maezawa et al., 2006; Hong et al., 2007). The death of MC65 neuroblastoma cells is the result of the intracellular accumulation of AβO which are immunoreactive to the oligomer-specific antibody A11 (Kayed et al., 2003). By selecting compounds that block this death, previously we showed that this method preferentially identifies those with good permeation properties and capable of interacting with AβO (Hong et al., 2007). Using this method, we screened amyloid ligands of various pharmacophore scaffolds and their derivatives, including those of styrylbenzene, benzothiazole, fluorene, stilbene, and biphenyl (Cai et al., 2007; Kung et al., 2001, 2002; Lee et al., 2003; Zhuang et al., 2001a,b, 2005). Results from selected compounds are summarized in Supplementary Table 1. While the application of IBOX (EC50 = 230 nM) and IMPY (EC50 = 260 nM), two thioflavin-like brain permeable amyloid ligands (Zhuang et al., 2001a; Kung et al., 2002), may be limited due to their narrow therapeutic window, the two fluorenes K01-162 and K01-186 (Lee et al., 2003) (Fig. 1A) are promising due to their highest MC65 protective effect and low toxicity. K01-162 showed full MC65 protection at 125 nM, an EC50 of 80 nM, and no cytotoxicity up to 50 µM (Supplementary Table 1). K01-186 has the highest MC65 protective effect (EC50 = 50 nM) and is a potent anti-oxidant (Bordwell et al., 1988; see below).
Corresponding to their cytoprotective effect, K01-162 and K01-186 reduced the level of intracellular AβO that are immunoreactive to the A11 antibody (Fig. 1B). A11 specifically recognizes the conformation of prefibrillary oligomers that may cause toxicity by increasing membrane conductance (Kayed et al., 2003, 2004). Western blot analysis using anti-Aβ3-8 antibody 6E10 showed that both K01-162 and K01-186 substantially reduced the levels of SDS-stable Aβ trimer and larger aggregates (Fig. 1C). The effect on Aβ dimer was not certain because it co-migrated with the putative CTFΔ31, a caspase cleavage product of C99 (Maezawa et al., 2006). The MC65-inactive fluorene analogs AD27 and K01-182 did not reduce the level of intracellular AβO. Thus, the ability of the compound to block MC65 death appears to correlate with its ability to reduce the level of intracellular AβO.
We further tested the ability of K01-162 and K01-186 to reduce intraneuronal AβO in primary neuronal cultures. Previously we showed that the intraneuronal accumulation of AβO composed of Aβ42 can be induced in APP-expressing primary cortical neurons or N2A neuroblastoma cells when treated with U18666A, a class 2 amphiphile that directly inhibits the function of NPC1 protein (Niemann-Pick type C disease protein 1) (Jin et al., 2004). The intraneuronal Aβ42 deposits are mainly localized to endosomes (Jin et al., 2004), consistent with the previously identified sites of Aβ oligomerization (Takahashi et al., 2004) and the location of intraneuronal Aβ in early AD brains and the brains of transgenic animals (Cataldo et al., 2004; Takahashi et al., 2004; LaFerla et al., 2007). As shown in Fig. 1D, intraneuronal A11-immunoreactive AβO were substantially attenuated by treatment with K01-162 and K01-186 (100nM to 10 µM). Western blot analysis using 6E10 demonstrated that U18666A-treated APP-expressing neurons contained two major SDS-stable AβO bands at 56 kDa and 80 kDa (Fig. 1E). The 56 kDa band corresponds in size with Aβ42 12-mer oligomers recently shown to correlate with cognitive deficits in Tg2576 mice, dubbed Aβ*56 (Lesne et al., 2006). Treatment with K01-162 and K01-186 substantially reduced the intensities of the two AβO bands. Due to the inherent toxicity of the U18666A, the neuroprotective effect of K01-162 and K01-186 in this culture system could not be assessed.
To provide a proof of principle that fluorenes thus selected have anti-Aβ properties in vivo, we tested K01-162 in a model of cerebral Aβ amyloidosis called 5xFAD mice (Oakley et al., 2006). In these mice, Aβ42 amyloid deposition begins by 2 months, especially in hippocampus and subiculum, and increases sharply between 2 and 4 months. We infused K01-162 intracerebroventricularly into 3-month-old 5xFAD mice for 2 weeks. This course did not cause apparent toxicity and significantly reduced the amyloid load in hippocampus to ~50% of the mock-treated level (Fig. 2A and B). While the immunostain-detectable amyloid load may roughly represent the insoluble fibrillar aggregates in the brain, we found that K01-162 infusion also reduced the level of RIPA buffer solubilized Aβ species that may better represent the smaller aggregates such as oligomeric Aβ (Oddo et al., 2006; Cheng et al., 2007), but did not affect the transgenic expression of the human amyloid-β precursor protein (hAPP) (Fig. 2C). Western blot showed a collectively ~50% reduction of the SDS-stable, mainly 9.7, 15.3, and 21.6 kDa oligomeric species, which may correspond to dimeric, trimeric, and tetrameric Aβ (Fig. 2C and D). Thus, K01-162 can reduce the brain amyloid burden that exists in both fibrillar and RIPA-soluble, non-fibrillar forms. In contrast, the similarly infused MC65-inactive fluorene K01-182 showed no such effect.
Previously we showed that Aβ fibrils and AβO may share common binding pockets for some amyloid ligands, therefore it is possible that the lead fluorenes, being able to inhibit AβO-induced cell death and reduce brain AβO levels, directly interact with AβO, in addition to binding to fibrils. To examine the binding of compounds to AβO, we employed our previously reported SPR method, which allows for the qualitative and quantitative measurements of interactions between the small-molecule compounds in the flow phase and the AβO immobilized on the sensor chip (Maezawa et al., 2008). The method is described in details in the Supplementary data. The advantage of this method includes the highly sensitive real-time detection of interactions and not requiring a labeling procedure which might change the property of the compounds. The SPR result showed that K01-162 and K01-186 bind directly to AβO with a KD of 19.0 µM and 20.4 µM, respectively, comparable to that of Congo red (CR) (Maezawa et al., 2008) (Fig. 3A). In contrast, the MC65-inactive fluorenes AD27 and K01-182 failed to show any binding.
What might be the consequence of the compound-AβO interaction? We hypothesize an immediate effect (seconds to minutes) to mask the conformation of AβO required for interaction with cellular targets (Townsend et al., 2006; Hong et al., 2007), and a delayed effect (hours to days) to destabilize the Aβ aggregates (Hong et al., 2007). To test the former, we determined the compounds’ ability to block the strong AβO binding to synapses, which was shown to induce aberrations in synapse composition, shape, and density, and was proposed as a molecular basis for loss of connectivity in AD (Lacor et al., 2004). This binding appears to depend on the oligomer conformation rather than the primary Aβ sequence (Lacor et al., 2004). Fig. 3B and C show that K01-162 and K01-186 blocked the synaptic binding activity of AβO after a preincubation for 5 min, an effect comparable to that of CR (Maezawa et al., 2008). This effect is not due to damage to synapses by the compounds, because prior incubation of the cultures with the compounds did not block subsequent AβO binding to synapses (data not shown). This result suggests that the interaction of K01-162 and K01-186 with AβO may provide an immediate synapto-protective effect. Significant effects were seen at 5–10 equiv. of compound to Aβ (molar ratio relative to Aβ monomeric peptide), which corresponds to a compound:peptide ratio of 0.3–0.6:1 (w/w). This is comparable to the effect of inositol stereoisomers, a group of previously reported promising anti-Aβ small molecules (McLaurin et al., 2000).
To test the delayed effect of fluorene compounds to destabilize AβO, we performed EPR spectroscopy (Hubbell et al., 2000) to observe the conformational dynamics of Aβ peptide as a function of oligomeric state and compound binding. EPR signals were analyzed from a paramagnetic nitroxide probe targeted specifically to the second residue of AβO (Supplementary Fig. 1). To alleviate dipolar broadening between nitroxide labels on neighboring peptides, the fraction of spin-labeled Aβ was diluted down to a 25 mol% fraction with native (unlabeled) Aβ. AβO thus made containing spin-labeled peptide were shown to have a comparable degree of toxicity to N2A cells (Hong et al., 2007) and a similar particle size distribution as the unlabeled AβO (data not shown). We incubated the spin-labeled AβO with candidate compounds for 1 h and examined the EPR spectra. As shown in Fig. 4, the EPR spectra of samples treated with K01-162 or K01-186 display additional line narrowing compared to the solvent control or the inactive compound K01-182. The decreased line width of spin-labeled AβO in the presence of K01-162 or K01-186 reflects increased disorder of the spin-labeled side chain at position 2 on Aβ. CR, known to disaggregate Aβ, also induces a similar increase in the molecular dynamics (Fig. 4). In contrast, the inactive compounds K01-182 (Fig. 4) and AD27 (data not shown) did not alter the spectrum of AβO. The increase in motion likely arises from either a more flexible backbone or an increase in global rotational dynamics, consistent with an increased tumbling rate of a monomeric peptide versus the oligomer species.
The above EPR results suggest that K01-162 and K01-186 destabilize AβO. Indeed, direct visualization of AβO by atomic force microscopy (AFM, Fig. 5) revealed that K01-162, K01-186 and CR treatment resulted in apparently much less (estimated 60% less) detectable AβO particles, likely due to the compound’s ability to depolymerize AβO into smaller units not detectable by AFM. However, due to the uneven distribution of the particles on the AFM mica, the number of detectable particles was not quantified. Instead, quantitation of the sizes of detectable Aβ particles clearly showed that K01-162 and K01-186 were able to block the progressive polymerization of preformed AβO to larger aggregates that were observed in mock-treated samples (Fig. 5B). There was also a trend that K01-162 and K01-186 reduced the size of AβO to a more homogeneous population, as indicated by the left shift and sharper peak of the size distribution curve (Fig. 5C). Furthermore, this size analysis likely underestimated the destabilizing effect of fluorenes on AβO, considering the ~60% less detectable particles in K01-162-and K01-186-treated samples as noted above. In contrast, the MC65-inactive fluorene AD27 showed no such effect. Similar results were also obtained when particles were visualized by electron microscopy (EM, Supplementary Fig. 2). The same conclusion reached by both AFM and EM reduces the possibility that our results were due to compound-altered binding of AβO particles to the AFM mica or EM grids. Taken together, these results indicate that K01-162 and K01-186 are able to destabilize and disaggregate AβO.
We found that a previously identified potent antioxidant has the same structure as K01-186 (identified as 2,7-(Me2N)2FlH2; Bordwell et al., 1988). The presence of two dimethylamino functions at C2 and C7 of its fluorene ring stabilizes the cation radical intermediate in the oxidative process and is responsible for its antioxidative effect. Confirming this effect, we found that K01-186 was able to inhibit the horse radish peroxidase (HRP)-catalyzed oxidation reaction (Mishra et al., 2006) with an IC50 around 500 nM, ten fold higher than the EC50 (50 nM) of its MC65 protective effect (Supplementary Fig. 3). K01-162 as well as a few inactive compounds did not show significant antioxidative effect, indicating that the antioxidative effect is not required for the MC65 protective effect of fluorenes.
Our approach was initiated based on the hypothesis that intraneuronal accumulation of Aβ and its aggregation might be the first step of the amyloid cascade, and that intraneuronal AβO are highly toxic (LaFerla et al., 2007). Using our unique MC65 model, we quickly selected compounds that reduce the level of intracellular AβO and block AβO-related cell death. Mounting evidence also indicates that extracellular soluble AβO substantially damage neurons (Walsh et al., 2002; Kayed et al., 2004; Klein et al., 2004; Lesne et al., 2006). Our data demonstrate that the two lead fluorenes also interact with soluble AβO and block their harmful binding to synapses (Lacor et al., 2004). Furthermore, they are able to destabilize AβO, which may alter the aggregation into lesser toxic or more degradable forms and may explain their ability to reduce the level of AβO in cultured neurons and in 5xFAD mouse brains. In addition, K01-186 is a potent antioxidant. Considering that oxidative damage induced by Aβ or non-Aβ causes may contribute causally to AD-related pathology (Lin and Beal, 2006), the dual benefits of anti-oxidation and anti-Aβ provided by K01-186 might be preferable for treating AD patients.
We provide the initial proof of principle that the fluorenes have an in vivo effect. A short 2-week intraventricular infusion resulted in ~50% reductions in the levels of RIPA-soluble AβO and histologically quantified Aβ load in 5xFAD mice, a line with perhaps the most robust production of Aβ42. RIPA-solubilized samples would contain membrane-associated intraneuronal and matrix-bound extracellular Aβ, in addition to soluble extracellular AβO. Similarly RIPA-solubilized brain samples from J20 (Cheng et al., 2007), ARC (expressing Arctic-mutant human APP) (Cheng et al., 2007), and 3xTg-AD (Oddo et al., 2006) mice contained the putative Aβ*56 that was first shown in Tg2576 mice to be closely linked to memory deficits (Lesne et al., 2006). Interestingly, although we detected Aβ*56 in U18666A-treated cultured cortical neurons, we did not detect Aβ*56 in the 3-month-old 5xFAD mice despite their high Aβ42 levels. Instead, the 5xFAD samples displayed smaller aggregates corresponding to dimers, trimers and tetramers (Fig. 2C), suggesting that either they are the native assemblies in 5xFAD brains or the SDS-stable subunits of larger AβO that are assembled differently from Aβ*56 and therefore SDS-unstable. In any event, we found that fluorene compound treatment reduced the cumulative levels of these small aggregates in 5xFAD mice (Fig. 2D) and the level of AβO including Aβ*56 in primary cortical neurons (Fig. 1E). This is important in light of the observation that therapeutic interventions that reduce Aβ fibrils at the cost of augmenting AβO could be harmful (Cheng et al., 2007).
Our approach also benefited from using small libraries of derivatives or analogs of known amyloid-binding compounds with good pharmacokinetics properties and brain bioavailability (Lee, 2002). K01-162 should have excellent brain uptake in addition to very specific binding to cerebral Aβ aggregates, inferred from previous studies of the fluorene 125I-C2f (Lee et al., 2003). 125I-C2f has almost the same structure as K01-162 except that the Br in K01-162 was replaced by radioactive I-125. We have also measured the oral bioavailability of K01-186 and found that at 1 h after an oral dose of 25 mg/kg to wild-type mice, the brain level of K01-186 was 5.9 µg/g of brain (Hong and Jin, unpublished data), indicating a robust brain uptake. Taken together with our in vivo treatment data, K01-162 and K01-186 are likely to be brain-active compounds. Experiments are underway in our laboratories to determine whether K01-162 and K01-186 applied orally are able to block AβO toxicity in vivo by electrophysiological and behavioral studies using APP transgenic mice.
As with protein mis-folding disorders in general, efforts to understand the mechanism of Aβ toxicity have been frustrated due to the inherent instability of the oligomeric species, which complicates studies due to sample inhomogeneity. This is apparent when AβO preparations are viewed under AFM and EM (Fig. 5 and Supplementary Fig. 2). Here we employed a novel approach using EPR spectroscopy of site-directed spin labels to evaluate compound binding to AβO and the resulting conformational changes in AβO. This method (Hubbell et al., 2000) allows us to examine Aβ structure and assembly from a diverse set of sample states under physiological conditions. We are able to observe changes in the conformational dynamics of Aβ induced by compound binding in a real-time manner (Fig. 4). Our success in this approach encourages future dissection of the conformational features of AβO by EPR analysis of spin labels targeted to other locations within the Aβ sequence. In this regard, inhibitor compounds such as K01-162 and K01-186 will be highly useful for elucidating the molecular properties inherent to the toxic Aβ conformation. Site-directed spin labeling experiments will also facilitate the testing of models for compound-Aβ interaction, and provide insights for future compound design.
We are grateful to Drs. Hank Kung and Mei-ping Kung for providing the tested compounds and Dr. Danielle Harvey for statistical assistance. This work was supported by grants from the Alzheimer Association (IIRG-06-26782), the California State Alzheimer’s Disease Research Fund, and the National Institute of Health (AG025500 and AG029246).
Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.neurobiolaging. 2008.09.019.
The compounds reported here have been the subject of a patent application. The University of California has entered into an agreement with a co-owning university to manage the patent application. Should the subject of the invention be commercialized and revenue generated, Dr. Jin would receive an inventor share of such revenue directly from the University of California as dictated by the UC Patent Policy.