A Microliter-Scale High-throughput Screening System with Quantum-Dot Nanoprobes for Amyloid-β Aggregation Inhibitors

The aggregation of amyloid β protein (Aβ) is a key step in the pathogenesis of Alzheimer’s disease (AD), and therefore inhibitory substances for Aβ aggregation may have preventive and/or therapeutic potential for AD. Here we report a novel microliter-scale high-throughput screening system for Aβ aggregation inhibitors based on fluorescence microscopy-imaging technology with quantum-dot Nanoprobes. This screening system could be analyzed with a 5-µl sample volume when a 1536-well plate was used, and the inhibitory activity could be estimated as half-maximal effective concentrations (EC50). We attempted to comprehensively screen Aβ aggregation inhibitors from 52 spices using this system to assess whether this novel screening system is actually useful for screening inhibitors. Screening results indicate that approximately 90% of the ethanolic extracts from the spices showed inhibitory activity for Aβ aggregation. Interestingly, spices belonging to the Lamiaceae, the mint family, showed significantly higher activity than the average of tested spices. Furthermore, we tried to isolate the main inhibitory compound from Satureja hortensis , summer savory, a member of the Lamiaceae, using this system, and revealed that the main active compound was rosmarinic acid. These results demonstrate that this novel microliter-scale high-throughput screening system could be applied to the actual screening of Aβ aggregation inhibitors. Since this system can analyze at a microscopic scale, it is likely that further minimization of the system would easily be possible such as protein microarray technology.


Introduction
Many neurodegenerative disorders such as Alzheimer's disease (AD), Parkinson's disease, prion disease, and Huntington's disease are associated with the aggregation and deposition of misfolded proteins, the amyloids [1,2]. Since aggregates containing oligomers and fibrils show toxicity towards neuronal cells, amyloid aggregation inhibitors may be key compounds in the regulation of these amyloid diseases. In general, inhibitory activity against amyloid aggregation is assessed by spectrophotometric assays using amyloid β protein (Aβ)-binding dyes (e.g., thioflavin-T (ThT) and Congo red [3,4,5]) or by direct observation of aggregates using transmission electronic microscopy (TEM) [6,7,8] and/or atomic force microscopy (AFM) [9,10]. However, dye-binding assays that use ThT and Congo red to evaluate these inhibitory effects could be prone to false positive effects because fluorescence intensities of these dyes may be influenced by the inner filter effects of contaminating compounds and/or the inhibitors themselves [11]. Moreover, there is a possibility of competition for binding between the dyes and inhibitors to amyloid fibrils. These problems are significant when screening novel active compounds. On the other hand, direct observation by TEM and AFM with fixation and washing processes is unsuitable for quantification and high-throughput analysis. Therefore, we have been attempting to develop a novel high-throughput screening system for amyloid aggregation inhibitors.
Recently we successfully developed real-time imaging and quantification of Aβ 42 aggregation using quantum-dot (QD)labeled Aβ 40 (QDAβ) [12]. In that study, we showed that QDAβ, which had a binding ratio (Aβ 40 /QD) of 6, was incorporated into Aβ fibrils with a similar efficiency as unlabeled Aβ 42 when 0.1-0.01% QDAβ was mixed with unlabeled Aβ 42 [12]. The time to reach the steady state was similar to that of intact Aβ 42 measured from turbidity measurements [13]. Moreover, there are only small amounts (0.1-0.01%) QDAβ relative to unlabeled Aβ 42 , suggesting that this imaging technology can monitor the physiological aggregation of intact Aβ 42 . Since the amount of aggregates decreased in the presence of anti-Aβ antibody, we proposed that this imaging technology can be applied to the microliter-scale screening of inhibitory substances for Aβ 42 aggregation [12].
To attain a microliter-scale and high-throughput screening system, in this study, we tried to optimize the observation methods and to develop a simple quantification method from fluorescence microscopic images. Furthermore, in order to confirm whether this microliter-scale high-throughput screening system can actually be used to screen Aβ aggregation inhibitors, we attempted to screen inhibitory substances for Aβ aggregation from 52 dried spices that include various plant species that are readily commercially available.

Methodology of the microliter-scale high-throughput screening system for Aβ aggregation inhibitors
In this study, we used a 1536-well plate. When a 96-well plate was used in a recent study [12], at least 40 µl of sample volume was necessary to prevent the sample from drying. The use of a 1536-well plate enabled the observation of aggregates using a 5-µl sample volume without the drying side-effect. To develop an image of Aβ aggregates, in this study, we used Qdot655 and a customized the fluorescence filter block whose excitation filter was adjusted from 425 nm to 470 nm to attenuate the inner filter effect by substances in samples. For example, maximum absorption of curcumin, which is a known Aβ aggregation inhibitor, is 426 nm [11,14], so that fluorescence intensity may be influenced if an excitation filter of 425 nm is used. Since Qdot655 has a higher extinction coefficient than Qdot525, which we used recently [12], over a wide range of excitation wavelengths, Qdot655 fluorescent light could be sufficiently detected by a general CCD camera, even when the excitation wavelength was 470 nm.
When 30 nM of QDAβ and 30 µM of Aβ 42 were incubated in a 1536-well plate at 37 °C for 24 h, their aggregates were visualized, as reported in our recent study [12] (Figure 1). First, we examined the effect of EtOH on Aβ aggregation because we used EtOH extracts of spices in the next screening step ( Figure 2). 40% EtOH immediately induced amorphous aggregates, which are not fibrils, of QDAβ and Aβ 42 , and the amorphous aggregates were eliminated by centrifugation at 10,000 xg for 2 min at 4 °C before incubation. Consequently, no fluorescence was observed in the sample in the presence of 40% EtOH (Figure 2, 40%). When the sample was incubated with 20% EtOH, it seemed that the aggregates were slightly reduced (Figure 2, 20%). The data showed that the aggregation was not affected by 10% EtOH (Figure 2, 10%), so we performed all experiments in the presence of 5% EtOH in this study.
Next, we explored a way to quantify the amount of aggregates from the fluorescence microscopy images. We had already reported that the volume of Aβ aggregates can be estimated from 3D-images by confocal fluorescence microscopy [12]. However, the time taken for 3D-image  acquisition is longer than that for 2D-image acquisition, and data size of the 3D-image is also larger than that of the 2Dimage because one 3D-image reconstruction requires several dozen 2D-images. Therefore, we considered a simple quantification method from one 2D-image. Before incubation, QDAβ molecules were dispersed in a sample solution, so that the fluorescence micrograph showed a uniform red color ( Figure 3A; left panel). After incubation, Aβ aggregates that were visualized by QDAβ were observed on the well bottom ( Figure 3A; right panel). This aggregation resulted inhomogeneous distribution of fluorescence intensity in micrographs ( Figure 3B). The histogram of fluorescence intensities of 10,000 pixels (100 × 100 pixels) prior to sample incubation was narrow but that after sample incubation was broad, and the standard deviation (SD) of post-incubation samples was larger than that of pre-incubation samples ( Figure  3C), suggesting that SD values correlated with the amount of Aβ aggregates. That is, as Aβ aggregation progressed, the variability of fluorescence intensities of each pixel increased, and standard deviation (SD) values also increased.
To confirm the correlation between the amount of Aβ aggregates and SD values, we attempted a concentrationdependent assay ( Figure 4). 30 nM QDAβ and various concentrations of Aβ 42 were incubated in a 1536-well plate and the images ( Figure 4A) were analyzed using ImageJ software (NIH). When the Aβ 42 concentration was low (10-25 µM), diffused QDAβ molecules were observed in the intervening space between Aβ aggregates, so that the intervening spaces remained red due to QDAβ (Fig. 4A). When Aβ 42 concentration was around 30 µM, almost all QDAβ molecules were inserted into Aβ 42 fibrils, so that the intervening spaces were darker than that of the low Aβ 42 concentrations (Fig. 4A). When Aβ 42 concentration was high (40-50 nM), accumulated Aβ aggregates at the well bottom were thick, so that out-of-focus aggregates were imaged in the fluorescence micrographs, causing a blurred image (Fig. 4A). The thickness of Aβ aggregates on the well bottom was about 50 µm when 100 µM Aβ 42 was incubated in a 96-well plate [12]. The depth of sample solutions in the 96-well plate and in the 1536-well plate is the same (2 mm), suggesting that the thickness of aggregates might be about 15 µm in the presence of 30 µM Aβ 42 . Since the depth of focus of the microscopic system used in this study was approximately 14 µm it is likely that the gradual decrease of SD values ( Figure 4B) over 30 µM of Aβ 42 concentration was caused by an increase of out-of-focus aggregates. Linear regression analysis showed that the SD values from the fluorescence micrographs increased in a concentrationdependent manner at less than 30 µM of Aβ 42 (R 2 = 0.93) ( Figure 4B). In this study, therefore, we screened using 30 µM Aβ 42 in the following steps. Furthermore, we confirmed whether time-dependent SD values increased as Aβ aggregation progressed. The micrographs showed time-dependent aggregation ( Figure 5A), and the SD values also increased in a time-dependent manner ( Figure 5B). The time-dependent graph ( Figure 5B) showed a typical kinetic curve for amyloid aggregation which consisted of time lag, growth, and steady state phases, similar to recent 3D volume data by confocal microscopy [12]. Although we have no conclusive evidence whether SD values increase in direct proportion to the amount of Aβ 42 aggregates, the concentration-( Figure 4B) and timedependent ( Figure 5B) SD data suggests that the SD values could be used as an approximate indicator of Aβ aggregates. Since the time-dependent data revealed that the aggregation reached a plateau around 24 h ( Figure 5B), incubation time was fixed at 24 h in the following screening steps.
On the basis of the above results, we set up the microliterscale high-throughput screening system as mentioned in the Experimental section and Figure 6. Simply put, various concentrations of candidate inhibitors were incubated with 30 nM QDAβ and 30 µM Aβ 42 in a 1536-well plate (Figures 6 and 7A) at 37 °C for 24 h, and SD values of fluorescence intensity in each micrograph were analyzed by ImageJ software. The halfmaximal effective concentration (EC 50 ) values of inhibitors could be estimated from the inhibition curves ( Figure 7B). Figure 7D shows the actual inhibition curves of well-known Aβ aggregation inhibitors that were determined from the SD values of fluorescence micrograph data ( Figure 7C). The EC 50 could be estimated from the inhibition curves using an EC 50 shift by the Prism global fitting program (GraphPad software).

Comprehensive screening of Aβ aggregation inhibitors from 52 spices
Since we confirmed that the microliter-scale high-throughput screening system can be applied to estimate the EC 50 of Aβ aggregation inhibitors, we then tried to actually screen the EtOH extracts of 52 dried spices belonging to 19 plant families ( Figure 8). The results revealed that many spices (about 90%) had inhibitory activity for Aβ aggregation when the extracts were added at a high concentration (~10 mg). On the other hand, there are five spices that did not show inhibitory activity (Figure 8), demonstrating that this high degree of inhibitory success was not caused by methodological problems of this screening system. That is, it is likely that the "not effective"  spices were comparable to negative controls of this screening system. The spices belonging to the Lamiaceae and Myrtaceae families showed significantly higher inhibitory activity against Aβ aggregation than other family members. Especially, thyme (Thymus vulgaris) (EC 50 = 0.035 ± 0.015 mg/ml), summer savory (Satureja hortensis) (EC 50 = 0.049 ± 0.056 mg/ml), and spearmint (Mentha spicata) (EC 50 = 0.018 ± 0.005 mg/ml), all belonging to the Lamiaceae (mint) family, affected Aβ aggregation at 0.01 mg/ml (Figure 8, gray cells on 0.01 mg/ml line). Although green pepper (Piper nigrum) also affected SD values at 0.01 mg/ml it is likely that the result was an error because the effect was not observed at 0.1 mg/ml of extract  concentration. All spices that did not show inhibitory activity ( Figure 8, poppy seed (Papaver somniferum), basil seed (Ocimum basilicum), red pepper (Capsium annuum), coriander (Coriandrum sativum), and cumin (Cuminum cyminum)) were spices originating from seeds while only red pepper contained fruit.
We then tried to isolate the main active substance from summer savory, which was one of the spices that showed the highest inhibitory activity, under the guidance of the inhibitory effect (EC 50 ) against Aβ aggregation determined by the microliter-scale high-throughput screening system (Figure 9). The ethanolic extract (32 g) of dried summer savory (500 g) was successively partitioned between CHCl 3 , AcOEt, n-BuOH and H 2 O ( Figure 9A). The AcOEt layer, the best active fraction, was applied to silica gel column chromatography, reverse phase column chromatography, gel filtration, and reverse phase HPLC to give a single compound (131 mg) ( Figure 9A), which had ESI-MS, NMR, and IR spectra identical to rosmarinic acid (RA). The inhibitory activities of the isolated RA and standard RA analyzed by the microliter-scale high-throughput screening system ( Figure 9B, top) and the ThT assay ( Figure  9B, bottom) were similar, supporting the conclusion that the isolated substance was RA.

Discussion
In this study, we successfully developed a novel microliterscale high-throughput screening system with QD for Aβ aggregation inhibitors ( Figure 6). The comprehensive screening of Aβ aggregation inhibitors for 52 spices using the screening system revealed that almost all spices had inhibitory activity for Aβ aggregation (Figure 8). Masuda et al. [15] reported that many polyphenols inhibit the aggregation of Aβ and α-synuclein, which is another amyloid. Polyphenols are natural substances present in many plants that are consistent with our result showing that almost all of the 52 spices tested had inhibitory activity. Furthermore, we demonstrated that the Lamiaceae family, except for basil seed, showed higher inhibitory activity against Aβ aggregation than all other plant families (Figure 8), and that RA was the main active substance of summer savory (Figure 9). These results suggest that this system could be applied to the actual screening of Aβ aggregation inhibitors.
RA was previously reported to inhibit Aβ aggregation in vitro [16] and Aβ deposition of RA-treated AD model transgenic mice decreased significantly in the brain [17]. Moreover, it was reported that RA prevented Aβ oligomerization and synaptic dysfunction by site-specific binding [18]. Interestingly, Salvia officinalis (sage) and Melissa officinalis (lemon balm), also belonging to the Lamiaceae family, had already been reported to have memory-improving properties in old European reference books hundreds of years ago [19]. It is possible that the effect involves the inhibitory activity of Aβ aggregation of RA, which is found in members of the Lamiaceae family [20]. We also found that some fractions, which were separate fractions in the isolation process from summer savory ( Figure  9), showed higher inhibitory activity than RA, but that the content was smaller than RA, thus, they were not major active Figure 8. Estimation of EC 50 values of EtOH extracts from 52 spices using the microliter-scale high-throughput screening system. Screening was applied to examine the EtOH extracts of dried spices. The 'black', 'gray', and 'white' cells indicate 'not inhibited (SD ≥ 80%)', 'partially inhibited (80% > SD > 20%)', and 'completely inhibited (20% ≥ SD)' wells, respectively. The percentage of SD was defined as SD values before and after incubation of control samples (0% and 100%, respectively). EC 50 values were estimated from dosedependent inhibition curves (n=3 separate experiments). Spices were aligned using the Angiosperm Phylogeny Group classification (APG III) [25]. doi: 10.1371/journal.pone.0072992.g008 fractions. Since this screening system can evaluate a small amount of sample, we are now trying to isolate the highly active compounds from the fractions (data to be published soon).
The EC 50 values that were determined by the microliter-scale high-throughput screening system (Figures 7 and 9) showed a tendency to be higher than the values that were determined by ThT fluorescent intensities [15,16,21]. As mentioned above, it has been reported that a dye-binding assay using ThT has the potential to show false positive effects such as inner filter effects [11]. In contrast, since this screening system adopts a longer excitation spectrum and quantification from variability data of fluorescence intensity, it is likely that the inner filter effects are smaller than those of the ThT assay.
Recently, potent amyloidogenicity and pathogenicity of Aβ 43 was reported [22]. The microliter-scale high-throughput screening system can be applied easily by changing unlabeled Aβ 42 peptide to Aβ 43 peptide. Furthermore, this system could be applied to screening inhibitors for the aggregation of other amyloid peptides such as α-cynuclein, huntingtin, and prion. In addition, it is likely that this principle can be used to screen inhibitors or promoters for polymerization of normal filamentforming protein such as actin and microtubules.
The toxicity of soluble oligomers is a trigger for neurodegenerative diseases [23,24], indicating that Aβ oligomerization inhibitors are also important for preventive and/or therapeutic agent. The growth of aggregates occurred after the nucleation step, i.e., the oligomerization step, suggesting that oligomerization inhibitors also affect the amount of Aβ aggregates in the same incubation period. In fact, it was recently demonstrated that RA and Myr blocked Aβ oligomerization [18]. Therefore, it is reasonable to assume that oligomerization inhibitors can be screened as a second screening after the first comprehensive screening of Aβ aggregation inhibitors using this microliter-scale highthroughput screening system. We have already proposed a quantification method for oligomerization using QDAβ in a recent report [12]. It is likely that the combination of these methods would be useful to screen inhibitory substances for each aggregation process.
Although this system could quantify Aβ aggregation with only a 5-µl sample volume when a 1536-well plate was used, further minimization is possible such as protein microarray technology because the microscopic sample size is sufficient for data analysis in this system. This minimization would allow for more high-throughput screening. In addition, it is also one of the biggest advantages that this system does not need any handling processes, such as sample transfer by pipetting, after aggregation. Amyloid aggregates are hydrophobic and stick to plastic surfaces, and therefore are easy to adhere to, including to micropipette tips, microtubes, and multi-well plates. This can be a cause of a serious experimental error in the quantification of aggregates.
Although many compounds have shown inhibitory activity against Aβ aggregation [15], the mechanism of inhibition remains unknown. Since this microliter-scale high-throughput screening system can be applied to comprehensive screening, the isolation of Aβ aggregation inhibitors from natural sources and chemical libraries may be accelerated from now on. It is expected that a quantitative structure-activity relationship study, which would use many types of inhibitors that can be identified by this screening system, would demonstrate an inhibition mechanism. Such information would be helpful in the molecular design of inhibitors in drug discovery. Our hope is that this novel technology may be a powerful tool for the analysis of protein aggregation/assembly.

Preparation of QDAβ
QDAβ was prepared according to our recent report [12] and a book (K. Tokuraku & T. Ikezu, Imaging of amyloid-β aggregation using a novel quantum dot nanoprobe and its advanced applications, Bionanoimaging-Insights into protein misfolding & aggregation, edited by Yuri Lyubchenko & Vladimir Uversky, Elsevier, in press). 10 µM QD-PEG-NH 2 was first reacted with 1 mM sulfo-EMCS in PBS for 1 h at 20 °C . After quenching and elimination of unreacted sulfo-EMSC, the QD-PEG-NH 2 -bound sulfo-EMCS was reacted with 100 µM CysAβ in PBS containing 5 mM EDTA for 1 h at 20 °C . The labeling ratio of Aβ to QD can be controlled by the initial concentration of the added CysAβ [12]. In this condition, the labeling ratio (Aβ/QD) was estimated as being approximately 6 [12]. The concentrations of QDAβ were determined at an absorbance of 504 nm according to instruction manual of Life Technologies.

Microliter-scale high-throughput screening system
Dried Aβ 42 films were completely dissolved in 1 mM DMSO by pipetting for at least 15 min with the tip of a micropipette pressed on the bottom of a microtube. QDAβ and the dissolved Aβ 42 were mixed in 1x PBS at final concentrations of 60 nM and 60 µM, respectively. 3 µl of the mixture of QDAβ and Aβ 42 and 3 µl of various concentrations of inhibitors or spice extracts, which were diluted in 1x PBS containing 10% EtOH, were mixed (final condition was various concentrations of inhibitors or spice extracts, 30 nM QDAβ, 30 µM Aβ 42 , 1x PBS, 5% EtOH, 3% DMSO), and centrifuged at 10,000 xg for 2 min at 4 °C to eliminate any insoluble dusts. 5 µl of the supernatants were transferred into each well of a 1536-well plate (782096, Greiner), and were tightly sealed with vinyl tapes to prevent the drying of sample solutions in wells. The 1536-well plate with sample solutions was centrifuged by a multi-well plate centrifuge (PlateSpin II, Kubota) at 1,530 xg for 5 min at room temperature to flatten the surface of sample solutions, and incubated at 37 °C for 24 h in an air incubator (SIB-35, Sansyo). The 1536-well plate was observed by an inverted fluorescence microscope (Diaphot 300, Nikon) equipped with a color CCD camera (DP72, Olympus). QDAβ was imaged using a customized Qdot655 filter set (XF305-1, Omega optical; excitation 425DF45, dichroic 475DCLP, emission 655DF20) in such a way that the excitation filter was replaced by another filter (Ex 450-490 nm, Nikon) in order to reduce the interfering QD excitation by components in spice extracts and by the inhibitory compound itself.

Estimation of EC 50 from fluorescence microscopy images
To estimate the half-maximal effective concentration (EC 50 ) that would inhibit Aβ aggregation, we used fluorescence micrograph data that was taken using a 4x objective lens with a customized Qdot655 filter set. Fluorescence intensities of 10,000 pixels (100 × 100 pixels: 280 × 280 µm) in the central region of each well and the SD values were measured by ImageJ software (NIH). The means and ±SDs of the mean values were calculated from three independent experiments. We defined EC 50 as the inhibitor concentration when the SD value showed half of the maximum (after incubation) and minimum (before incubation) SDs of control sample without inhibitor. In particular, the inhibition curves (e.g. Figure 1G) were analyzed with Prism (GraphPad software) using an EC 50 shift by global fitting (Asymmetric sigmoidal, 5 parameter logistic).

Estimation of EC 50 from ThT fluorescence intensities
EC 50 for Aβ aggregation by ThT fluorescence was determined according to the method of Levine [4] modified in our laboratory. ThT (202-01002, Wako Pure Chemical Industries) was dissolved in water at 100 µM for 1 day at room temperature under constant stirring and shielded from light. The 100 µM ThT solution was diluted immediately before use in 50 mM glycine/KOH buffer (pH 8.5) to a final concentration of 5 µM. 30 µM Aβ 42 and various concentrations of inhibitors were incubated in 1x PBS containing 5% EtOH and 3% DMSO for 24 h at 37°C. 10 µl of the incubated samples were added to 190 µl of 5 µM ThT solution, and the fluorescence intensity was measured at 490 nm after excitation at 455 nm in micro fluorimeter cells (FM20B, GL Science) using a fluorescence spectrophotometer (F-4500, Hitachi). EC 50 was estimated from an inhibition curve using the EC 50 shift by Prism global fitting software.

Preparation of EtOH extracts from 52 dried spices
One g of each dried spice was extracted with 5 ml of EtOH at room temperature for 7 days. The EtOH solution was filtered and concentrated in vacuo.