Halogenation Generates Effective Modulators of Amyloid-Beta Aggregation and Neurotoxicity

Halogenation of organic compounds plays diverse roles in biochemistry, including selective chemical modification of proteins and improved oral absorption/blood-brain barrier permeability of drug candidates. Moreover, halogenation of aromatic molecules greatly affects aromatic interaction-mediated self-assembly processes, including amyloid fibril formation. Perturbation of the aromatic interaction caused by halogenation of peptide building blocks is known to affect the morphology and other physical properties of the fibrillar structure. Consequently, in this article, we investigated the ability of halogenated ligands to modulate the self-assembly of amyloidogenic peptide/protein. As a model system, we chose amyloid-beta peptide (Aβ), which is implicated in Alzheimer’s disease, and a novel modulator of Aβ aggregation, erythrosine B (ERB). Considering that four halogen atoms are attached to the xanthene benzoate group in ERB, we hypothesized that halogenation of the xanthene benzoate plays a critical role in modulating Aβ aggregation and cytotoxicity. Therefore, we evaluated the modulating capacities of four ERB analogs containing different types and numbers of halogen atoms as well as fluorescein as a negative control. We found that fluorescein is not an effective modulator of Aβ aggregation and cytotoxicity. However, halogenation of either the xanthenes or benzoate ring of fluorescein substantially enhanced the inhibitory capacity on Aβ aggregation. Such Aβ aggregation inhibition by ERB analogs except rose bengal correlated well to the inhibition of Aβ cytotoxicity. To our knowledge, this is the first report demonstrating that halogenation of aromatic rings substantially enhance inhibitory capacities of small molecules on Aβ-associated neurotoxicity via Aβ aggregation modulation.


Introduction
Halogenation has been widely used to provide organic compounds including biomolecules with new properties. Introduction of aryl halides into proteins allows chemical modification via versatile palladium catalyzed cross-coupling reactions with terminal alkene or alkyne reaction partners [1,2], and facilitates monitoring structural changes of protein [3,4]. Halogen groups are often inserted during hit-to-lead or leadto-drug conversions for several reasons, including enhanced antagonistic/agnostic effects due to improved oral absorption/ blood-brain barrier permeability [5]. Furthermore, it was reported that halogenation of aromatic molecules greatly affects aromatic interaction-mediated self-assembly processes [6]. Aromatic interaction plays an important role in a broad spectrum of molecular self-assemblies [3,7,8,9]. In particular, aromatic interaction is considered one of critical contributors to forming cross-stacked b-sheet structure, so called, amyloid fibrillar structure [10,11]. Planar aromatic interaction stabilizes the fibrillar structure and determines the direction and orientation of amyloid fibrils [12,13]. Therefore, perturbation of the aromatic interaction caused by halogenation of aromatic building block affects the morphology and physical properties of the fibrillar structure [3].
Herein, we have investigated whether halogenation of ligands can also affect self-assembly of amyloid-beta peptide (Ab), which is implicated in Alzheimer's disease (AD). A pathological hallmark of AD is the accumulation of insoluble protein aggregates, composed primarily of fibrillar Ab aggregates. According to the revised amyloid-cascade hypothesis, certain types of soluble Ab oligomers and protofibrils are more toxic than Ab fibrils and correlate well with dementia [14,15,16,17]. Therefore, modulation of Ab aggregation using small molecules is considered a promising way to eliminate Ab associated toxicity [3,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32]. We recently reported that red food dye erythrosine B (ERB) is a novel modulator of Ab-aggregation in vitro and Ab neurotoxicity [33]. The good biocompatibility and possibility of systemic administration make ERB an attractive inhibitor of Ab neurotoxicity [34,35]. Considering that ERB has multiple aromatic rings attached to four electronegative halogen atoms (Figure 1), we hypothesize that the modulatory capacity of ERB on Ab aggregation is attributed to halogen atoms. In order to validate our hypothesis that halogen atoms are key chemical structures for Ab aggregation modulation, we evaluated the modulating capacities of four ERB congeners containing different type and number of halogen atoms, eosin Y (EOY), eosin B (EOB), rose bengal (ROB), and phloxine B (PHB) (Figure 1). As a negative control, we also evaluated fluorescein (FLN), which has the same xanthene benzoate backbone as ERB but lacks a halogen atom. If halogenation of aromatic rings is indeed effective in modulating Ab aggregation and cytotoxicity, it will enhance our understanding of molecular mechanism of amyloid formation and facilitate discovery and design of a new series of halogenated small molecule modulators of amyloidogenic peptides/proteins.

Ab Aggregation
Ab40 stock was prepared as described previously [33,36] except the pretreatment using hexafluoroisopropanol (HFIP). It has been reported that HFIP increases the a-helix content of a protein and is a strong disaggregating solvent of Ab [37,38]. Lyophilized Ab40 was dissolved in 100% HFIP (1 mM) and incubated at room temperature for 2 hours. HFIP was evaporated under a constant stream of nitrogen, and the peptide was reconstituted in phosphate buffered saline (PBS) solution (10 mM NaH 2 PO 4 and 150 mM NaCl, pH 7.4) to a concentration of 50 mM. If needed, the HFIP treated peptide was dissolved in 100 mM NaOH (2 mM Ab) prior to dilution in PBS. Erythrosine B, eosin Y, eosin B, rose bengal, phloxine B, and fluorescein were dissolved in PBS. Concentrated dye stock solutions were added to the peptide solutions. The Ab40 peptide samples were incubated at 37uC in the absence or in the presence of the dye.
Thioflavin T (ThT) Assay 5 mL of Ab sample (50 mM) was dissolved in 250 mL of ThT (10 mM). Fluorescence was measured in 96-well microtiter plates (Fisher Scientific, Pittsburgh, PA) using a Synergy 4 UV-Vis/ fluorescence multi-mode microplate reader (Biotek, VT) with an excitation and emission wavelength of 438 nm and 485 nm, respectively.

Transmission Electron Microscopy (TEM)
TEM was performed as reported previously [33,36]. Ab samples (10 mL of 50 mM Ab) were placed on 200 mesh formvar coated/copper grids, absorbed for 1 minute, and blotted dry with filter paper. Grids were then negatively stained with 2% uranyl acetate solution, blotted dry, and then inspected with a JEOL 1010 Transmission Electron Microscope operated at 60 kV.

Dot Blotting
Dot blotting was performed as reported previously [33,36]. 2 mL Ab samples were spotted onto nitrocellulose membranes and were dried at room temperature. A solution of 0.1% Tween 20 in Tris-buffered saline (TBS-T) solution (0.1% Tween 20, 20 mM Tris, 150 mM NaCl, pH 7.4) was prepared. Each nitrocellulose membrane was blocked at room temperature for 1 hour (5% milk TBS-T) and washed with TBS-T. Each membrane was then incubated with antibody (HRP-conjugated 4G8, A11, or OC antibody) in 0.5% milk TBS-T for 1 hour at room temperature and washed with TBS-T. After immuno-staining with HRPconjugated 4G8, the membranes were coated with ECL advance detection agent (based on manufacturer specifications) and visualized. Alternatively, all other membranes were incubated with HRP-conjugated IgG in 0.5% milk TBS-T for 1 hour and washed with TBS-T. Signal detection was performed as aforementioned using the ECL Advance Detection kit and was visualized using a Biospectrum imaging system (UVP, Upland, CA). HRP-conjugated 4G8 and OC were applied at a 1:25000 dilution while A11 and HRP-conjugated IgG were applied at a 1:10000 dilution.

MTT Reduction Assay
MTT reduction assay was performed as reported previously [33,36]. SH-SY5Y cells were cultured in a humidified 5% CO 2 / air incubator at 37uC in DMEM/F 12:1:1 containing 10% fetal bovine serum and 1% penicillin-streptomycin. 20000 to 25000 cells were seeded into each well of a 96-well microtiter plate (BD, Franklin Lakes, NJ) and allowed to acclimate for 3 days. 10 mL of Ab sample was added to each well and incubated for 2 days. The cells were washed by replacing the culture media with fresh media and incubating for 1 hour. The wash media was replaced with fresh media. 10 mL of MTT was added to each well and incubated in the dark for 6 hours at 37uC. After incubation, reduced MTT was dissolved with 200 mL of dimethylsulfoxide (DMSO). After reduced MTT dissolution, the absorbance was measured at 506 nm using a Synergy 4 UV-Vis/fluorescence multi-mode microplate reader (Biotek, VT).

Circular Dichroism (CD)
CD analysis of Ab samples was performed as described previously [39,40]. Ab sample was diluted 1:10 using double distilled water. Samples were measured using a Jasco J710 spectropolarimeter with a 1 mm path length. The reported spectrum for each sample was the average of at least 5 measurements and the background was subtracted using appropriate controls. In case of samples containing any dye, the background spectra were obtained using controls containing only dye at the same concentration.

Ab Binding Assay
The binding of ERB, EOY, ROB, PHB, and FLN to Ab40 was assessed using modified assays based on emission fluorescence quenching techniques described in the literature [41,42,43,44]. The concentration of each of the dyes was fixed at 20 mM. In order to evaluate fluorescence quenching of the dye upon binding to Ab40, Ab40 was mixed with the dye in a final concentration of 0 to 25 mM in citrate buffer at pH 4.5. The excitation wavelengths used are as follows: ERB -317 nm, EOY -480 nm, ROB -510 nm, PHB -500 nm, and FLN -432 nm. The emission wavelengths where the data were collected are as follows: ERB -548 nm, EOY -536 nm, ROB -565 nm, PHB -555 nm, and FLN -512 nm. With FLN, fluorescence quenching was also investigated due to binding to bovine serum albumin (BSA -New England Biolabs, Ipswich, MA) by mixing with FLN in a final concentration of 0 to 25 mM BSA in citrate buffer at pH 4.5. Where appropriate, the dissociation constant, K d , was determined using the non-linear regression curve fitting to Eq. 1 shown below. In Eq. 1, n is the number of binding sites, and [D] is the molar concentration of free dye. We assessed the binding of EOB to Ab40 and BSA, using an absorbance technique described in the literature based on the observation that upon protein binding [45], the absorbance maximum of EOB shifts from 514 to 53065 nm. The concentration of EOB was fixed at 20 mM. Ab40 and BSA concentrations were varied from 0 to 60 mM and 0 to 25 mM, respectively, and the absorbance was measured at 530 nm. Citrate buffer at pH 4.5 was also used for the EOB binding assay.

ERB, EOY, and PHB Substantially Inhibit Ab-Associated Cytotoxicity
In order to evaluate the modulation capability of ERB and its analogs (EOY, EOB, PHB, and ROB), we employed the widelyused MTT reduction assay [16,29,33,36,46,47]. Ab aggregates were prepared by incubating Ab monomers with or without 3x ERB analog. In the absence of any ERB analog, Ab aggregation was monitored by ThT fluorescence assay. The ThT fluorescence of Ab aggregates started to increase at day 4 and reached the plateau at day 6 ( Figure 2A), indicating that Ab protofibrils and fibrils were primarily formed from day 4. In order to evaluate cytotoxicity of Ab aggregates containing Ab intermediates, we chose Ab samples incubated for 5 days in the absence or presence of 3x ERB analog. The preformed Ab aggregates were then administered to neuroblastoma SH-SY5Y cells, and cell viability was determined by MTT reduction ( Figure 2B). We determined whether Ab monomer or ERB analog is cytotoxic to neuroblastoma SH-SY5Y cells, and the results are shown in Figure 2B. Ab monomers (5 mM) caused a mild reduction (11%) in the cell viability. All ERB analogs (15 mM) except ROB also caused only mild reduction in the cell viability ranging from 0 to 8%. However, 3x ROB substantially reduced the cell viability (34%). ROB has been tested to ablate certain types of cancer cells including melanoma [48,49], and so it is not surprising that ROB is cytotoxic to SH-SY5Y cells.
Next, we determined the cytotoxicity of Ab monomers incubated with or without ERB analog for 5 days, and the results are shown in Figure 2B. 5 mM of Ab aggregates without any ERB analog (Ab control) substantially reduced the cell viability to 63%. Co-incubation of Ab monomers in the presence of 3x EOB (15 mM) resulted in an SH-SY5Y cell viability was 65%, which is not significantly different from that of Ab control. However, coincubation of Ab monomer with 3x ERB, EOY, or PHB significantly increased the cell viability (around 21%). In the presence of 3x ROB, cell viability was 70%, which is only 7% higher than that of the Ab control. The MTT reduction assay results clearly indicate that 3x ERB, EOY, and PHB can substantially inhibit Ab-associated cytotoxicity but 3x EOB cannot. The Ab monomers incubated with 3x ROB (15 mM) led to a substantial reduction in the cell viability (30%). However, since 3x ROB alone (no Ab) was intrinsically toxic and led to a similar reduction in cell viability (34%), it is difficult to gauge the effect 3x ROB co-incubation had on Ab-induced cytotoxicity. In order to clarify this, we repeated the MTT cell viability assay, this time comparing the results obtained using 2.5 mM and 5 mM Ab, both with corresponding concentrations of 3x ROB (7.5 mM and 15 mM, respectively - Figure S1; Panels A and B). Since the ThT fluorescence of the Ab aggregates reach a plateau at day 6, the Ab aggregates in day 3 were used as Ab intermediate controls. When 5 mM Ab and 15 mM ROB was used, we again observed a substantial reduction in cell viability upon the addition of 3x ROB alone (P,0.05) and Ab intermediate controls compared to Ab monomer and PBS samples ( Figure S1; Panel A). However, when concentrations of 2.5 mM Ab and 7.5 mM ROB were applied to the cells, the intrinsic cytotoxicity of ROB alone (no Ab) was greatly reduced to approximately the level of the Ab monomer control ( Figure S1; Panel B). These results allowed us to interpret the true effect ROB had on Ab-induced toxicity. Similar to EOB, co-incubation of Ab monomers with 3x ROB for 3 days did not significantly alleviate the Ab-associated cytotoxicity displayed by the Ab intermediate control (P.0.05). Next, in order to investigate the effect that dye binding to Ab had on Ab-associated toxicity, Ab intermediates from day 3 of aggregation were mixed with 3x ROB and immediately added to the SH-SY5Y cells. As with the ROB co-incubation, the results showed that ROB binding to Ab did not alleviate the associated toxicity (P.0.05) ( Figure S1). In addition, since the Ab intermediates mixed with 3x ROB immediately prior to addition to the cells showed similar cell viability to the Ab intermediate control, we concluded that the intrinsic toxicity of ROB and Ab are not additive.
It should be noted that careful execution of the MTT reduction assay and interpretation of the results is required due to several factors. The first potential issue is that of Ab-induced expedited exocytosis of the reduced MTT. Several reports showed that Ab aggregates can export the reduced MTT and so promote the crystalline form of the reduced MTT deposit on the cell surface leading to a reduced MTT uptake [50,51,52]. In our previous studies, there was a good correlation between a MTT reduction and other viability assay based on Alamar blue reduction [36]. Therefore, we considered the MTT reduction assay is a valid viability assay on the cell line and Ab preparation method used in this study. The second issue relates to potential interference effects that the dyes investigated in this study might have on the final results obtained from the cell viability MTT assay (itself a colorbased test). In order to minimize this potential interference by removing the dyes prior to reading the MTT signal, all viability assays incorporated thorough washing steps, as detailed in the Methods section. To validate the washing steps conducted, the fraction of each original dye amount remaining in the culture plate wells after thoroughly washing the cells using the MTT protocol was quantified. The results showed that less than 3% of the original dye amounts remained in the wells after washing (Table  S1). Next, we quantified the interference effect these residuals might have on the final MTT absorbance. Our results showed that the interference was less than 5% for all dyes (Table S1), which is consistent with the intrinsic uncertainty of the MTT assay (4 to 6%) in Figure 2B and Figure S1, indicating that the dyes do not cause significant spectral interference in the MTT assays.
By correlating the chemical structures of ERB analogs and their inhibitory capacities on Ab cytotoxicity, we deduced the following. First, EOY, which contains four bromine atoms in the same locations as the four iodine atoms in ERB, exhibited similar inhibitory capacities on Ab cytotoxicity as ERB. However, EOB, which contains two nitro groups in the place of the two bromine atoms in the xanthene group of EOY, did not show any significant inhibitory capacity on Ab cytotoxicity. Therefore, these findings clearly indicate that either bromine or iodine atoms in the two positions of xanthene group are critical for Ab cytotoxicity inhibition. Second, PHB, which contains four extra chlorine atoms in the benzoate ring structure present in EOY, exhibits significant inhibitory capacities on Ab cytotoxicity (similar to EOY). The third conclusion we made was in regards to ROB, which did not eliminate Ab-associated cytotoxicity. ROB differs from ERB in that it is outfitted with four extra chlorine atoms in the benzoate ring and differs from PHB in that the bromine atoms on the xanthene group are replaced with iodine. The ROB results clearly indicate that not only the presence, but also the specific position of the halogenation, are important in determining the potency in inhibiting Ab-cytotoxicity.

Ab Monomers Aggregate to Form Prefibrillar and Fibrillar Aggregates
In order to determine whether Ab cytotoxicity inhibition by ERB analogs is associated with Ab aggregation modulation, we characterized the Ab aggregates formed in the absence or presence of each ERB analog using CD, TEM, and dot-blot assays. CD analysis has been widely used to monitor secondary structure changes of proteins [53,54,55,56]. The CD spectrum of Ab monomer did not exhibit any spectral feature of a-helix and bsheet, but showed typical features of dominantly disordered structure ( Figure 3A). The CD spectrum of Ab aggregates at day 5 exhibited the typical signatures of b-sheet structure, including a minimum at 217 nm ( Figure 3A), which indicate that disordered Ab monomers aggregated into b-sheet rich fibrillar aggregates. The TEM image of Ab monomers incubated for 5 days also clearly show the existence of the Ab aggregates consisting of protofibrils and short fibrils (Figure 4; Panel Ab only). Recently, dot-blotting with Ab-specific antibodies was widely used to detect the spectrum of Ab aggregates with different conformations [16,27,37,57,58,59,60]. OC is a polyclonal antibody that reacts with neurotoxic fibrillar oligomers, protofibrils and fibrils [16,58]. It was shown that Ab-associated toxicity could be eliminated by reducing the OC-reactive species [16]. Dot-blot assay using the OC antibody confirmed the existence of fibrillar structure at day 5 ( Figure 5; Panel OC). 4G8 is an Ab-sequence-specific monoclonal antibody [61,62,63,64] of which epitope is known to be residues 17 to 24 of Ab. During transition from monomers to fibrils, bsheet stacking buries the 4G8 epitope and ultimately limits 4G8 antibody access to the epitope leading to a significant reduction in the 4G8 reactivity [33,36,65]. Therefore, the reduction in 4G8 reactivity of Ab aggregates at days 5 and 6 can be attributed to the formation of fibrils and the lateral fibril stacking ( Figure 5; Panel 4G8). A11 is a polyclonal antibody that reacts with disordered prefibrillar aggregates [16]. The weak A11-reactivity of the Ab aggregates at day 5 indicate that content of disordered prefibrillar Ab aggregates was low ( Figure S2). Therefore, the CD, TEM, and dot-blot results using Ab-specific antibodies clearly show that the Ab aggregates at day 5 mainly consist of fibrillar aggregates including protofibrils and short fibrils.

EOB Does Not Modulate Ab Aggregation, but PHB Substantially Inhibits Ab Aggregation
Next, we characterized the Ab aggregates formed in the presence of 3x or 10x EOB. The CD spectrum of Ab aggregates formed with EOB exhibits dominant b-sheet structure, possibly fibrillar structures, similar to that of Ab control ( Figure 3A). The TEM images also show that the EOB-induced Ab aggregates have protofibrils and short fibrils similar to the Ab control (Figure 4; Panels EOB and Ab only). Furthermore, the EOB-induced Ab aggregates exhibit immuno-reactivity against OC-, 4G8-, and A11-antibodies similar to those of the Ab control from days 0 to 6 ( Figure 5; Figure S2). The CD, TEM, and dot-blot assay results clearly indicate that the co-incubation of EOB with Ab monomer does not substantially affect Ab aggregation process, which is consistent with the MTT reduction results showing that the cytotoxicity of the EOB-induced Ab aggregates was comparable to that of Ab control ( Figure 2B). These findings indicate that addition of two nitro groups and two bromine atoms to xanthene benzoate does not enhance modulatory capacity on Ab aggregation and cytotoxicity. However, considering the possibility of negative effects of two nitro groups on the modulatory capacity of halogenated xanthene benzoates, we also tested the other xanthene benzoate derivatives which contain only halogen atoms.
In case of PHB, the CD, TEM, and dot-blot assay results clearly indicate that co-incubation of Ab monomer with PHB significantly inhibits the Ab aggregation process (Figures 3A, 4, and 5). First, the CD spectrum of the Ab monomers co-incubated with PHB for 5 days do not show any typical features of a-helical and b-sheet structure strongly indicating that the PHB-induced Ab species has the disordered structure ( Figure 3A). In the TEM image of the PHB-induced Ab species, no Ab aggregates were observed ( Figure  4; Panel PHB) indicating no large molecular weight aggregates are present in the Ab sample. Since no aggregates were detected in the TEM image, the dot-blot assays using fibrillar or disordered oligomer-specific antibodies (OC-or A11-antibodies) were employed to monitor formation of Ab oligomers. The Ab monomers  Figure S2). Therefore, the TEM, CD, and dot-blot assay results strongly support the idea that co-incubation of PHB significant inhibits formation of any Ab oligomers/higher molecular weight aggregates, but allows maintaining Ab monomer-like structural features. Considering that Ab monomer is known to be a non-toxic species [16,33,36], the substantial reduction of Abassociated cytotoxicity by co-incubating Ab monomer with PHB can be attributed to the Ab monomer-like structure of the PHBinduced Ab species.

EOY, ERB, and ROB Substantially Inhibit Fibrillar Structure Formation
We then characterized the Ab aggregates formed in the presence of 3x or 10x EOY, ERB, or ROB. The three CD spectra of the Ab aggregates formed with one of the three ERB congeners (10x EOY, ERB, and ROB) were almost overlapped ( Figure 3B), indicating that the secondary structure contents of the Ab aggregates are similar. The negative ellipticity value over all ranges of wavelength and the strong negative ellipticity values below 200 nm indicate the typical features of denatured proteins [66] or disordered Ab aggregates induced by small molecules [16,21]. Therefore, the CD analysis results support the idea that the three Ab aggregates formed with EOY, ERB, and ROB have an increased disordered structure content but a decreased b-sheet structure (possibly fibrillar structure) compared to Ab control. However, the overlapped CD spectra of the Ab samples with the three 10x dyes make it difficult to determine relative Abaggregation modulating capacities of the three dyes. Therefore, the CD spectrum of the Ab aggregates formed with a lower concentration (3x) of EOY, ERB, or ROB was also obtained  ( Figure S3). The estimated b-sheet content, possibly fibrillar structure, of the Ab samples with the three dyes based on the ellipticity value around 217 nm is in descending order of EOY, ERB and ROB. The TEM images of the three Ab aggregates formed with EOY, ERB, and ROB also show that the morphology of the three Ab aggregates are quite different from that of Ab control (Figure 4). The EOY-induced Ab aggregates are primarily small protofibrils in the length of 20 to 40 nm and a small portion of ,100 nm straight protofibrils (Figure 4; Panel EOY), whereas the Ab control mainly consisted of protofibrils and fibrils in the length of.300 nm (Figure 4; Panel Ab only). The ERB-induced Ab aggregates are curvilinear aggregates protofibrils, suggesting that the disordered structure content is higher than that of the Ab control (Figure 4; Panel ERB). The ROB-induced Ab aggregates also appeared as curvilinear protofibrils, but are thinner than the ERB-induced Ab aggregates (Figure 4; Panel ROB). Dot-blot assays using the OC and A11 antibodies were employed to estimate the relative amount of fibrillar and prefibrillar aggregates in the Ab samples. At day 5, the EOY-, ERB-, and ROB-induced Ab aggregates were in descending order of OC-reactivity ( Figure  5; Panel OC), which is quite consistent with the trend found in the CD analysis ( Figure S3). In contrast, the ROB-, ERB-, and EOYinduced Ab aggregates were in the descending order of A11reactivity ( Figure S2). Since the ROB-induced Ab aggregates exhibit very high A11-reactivity, we investigated whether there was any spectral interference of all ERB analogs with the dot-blot assay using the A11 antibody. The ERB congeners alone as well as the A11-reactive Ab aggregates were spotted to a nitrocellulose membrane and then the A11-reactivity of the samples was determined. Only ROB exhibits a significant A11-reactivity comparable to those of Ab samples ( Figure S4). Therefore, caution should be taken to interpret A11-reactivity of Ab samples containing ROB. None of the ERB congeners exhibit a significant immuno-reactivity against the OC and 4G8 antibodies (data not shown). The decrease in the OC-reactivity of the ERB analogs can be directly interpreted as a decrease in the fibrillar structure content, but the increase in the A11-reactivity of the ROB-induced should not be interpreted as an increase in the prefibrillar content.
For all three ERB congeners (EOY, ERB and ROB), the CD spectra, TEM images, and dot-blot assay using OC-antibody clearly indicate that there was a substantial of reduction in the fibrillar structure. Combined with the MTT reduction assay results (Figure 2B), such a reduction in the fibrillar structure can be attributed to a reduction in the Ab-associated cytotoxicity for EOY and ERB. Although the A11-reactivity of the ROB-induced Ab aggregates is greater than that of the Ab control, the A11reactivity is most likely overestimated. It is also interesting to note that even though ROB did not reduce Ab-associated cytotoxicity in the MTT assay, these results show that it is clearly a potent inhibitor of the Ab-aggregation.

FLN Does Not Effectively Modulate Ab Aggregation and Cytotoxicity
Investigating the modulatory capacities of ERB congeners on Ab cytotoxicity and aggregation reveals that even a subtle change in their chemical structure from the ERB structural template can affect their modulatory capacities. In order to further validate our hypothesis that the modulatory capacities of the ERB congeners are related with the presence of halogen atoms, we also evaluated the modulatory capacities of FLN as a negative control without any halogen atoms (Figure 1). The CD spectrum of the FLNinduced Ab aggregates clearly exhibits the typical features of bsheet rich structure ( Figure 6A). The TEM image of the FLNinduced Ab aggregates also indicates that protofibrils and fibrils are dominant species similar to the Ab control ( Figure 6B). Furthermore, the OC-reactivity of the Ab aggregates formed with FLN at days 5 and 6 are very comparable to those of the Ab control ( Figure 6C), indicating that the FLN-induced Ab aggregates had fibrillar aggregates as much as the Ab control. The 4G8-reactivity of the FLN-induced Ab aggregates with FLN remained unchanged up to day 7, whereas the 4G8-reactivity of the Ab control dropped at day 5. Such a slightly higher 4G8reactivity of the FLN-induced Ab aggregates at day 5 is likely because the FLN-induced fibrils are not laterally stacked and so allow the 4G8 binding to its epitope better than the Ab control. The CD, TEM, and dot-blot assay results conclusively demonstrate that FLN does not modulate the Ab aggregation as much as EOY, ERB, or ROB.
Next, we investigated whether FLN affects the Ab-associated cytotoxicity. Similar to the ERB analogs, Ab monomers were incubated in the absence of or presence of FLN for 5 days, and the resulting aggregates were subjected to the MTT reduction assay. The viability of the SH-SY5Y cells treated with the Ab control (5 mM) was 66% ( Figure 6D). Co-incubation of the Ab monomer with FLN led to a small increase in the cell viability (6%) ( Figure  6D), but the difference was only marginally significant (P = 0.013), while ERB, EOY, and PHB led to a substantial increase in the cell viability (P,0.001). The MTT assay results indicate that FLN did not substantially eliminate Ab cytotoxicity, which is consistent with the fact that FLN did not modulate Ab aggregation.

Halogenation of Xanthene Benzoate Generates Efficient Binders of Ab
Having discovered from the CD, TEM, and dot-blotting results that ROB, PHB, ERB, and EOY (but not EOB and FLN) are potent inhibitors of Ab aggregation, we then investigated possible correlations between these inhibition results and the binding affinity of the dyes to Ab. Dissociation constant (K d ) values and the number of binding sites were calculated for ROB, PHB, ERB, and EOY using fluorescence quenching of 20 mM concentrations of the dyes upon binding to Ab ( Table 1). The results showed that EOY most strongly binds Ab among the dyes used in this study. Intriguingly, the FLN (negative control lacking halogen atoms) quenching results showed that FLN is an exceptionally weak binder of Ab with less than 3% of the dye bound even in the presence of an excess molar concentration of 25 mM Ab ( Figure  S5; Panel A). In order to maintain consistency with the other five small molecules, our first preference was to employ a similar fluorescence quenching technique to assess the binding of 20 mM EOB (analog of EOY with replacement of the two bromine atoms close to benzoate group in EOY with two nitro groups) to Ab40. Despite varying reports in the literature about the fluorescence of the EOB molecule [41,42,43,44] and trying various solvents and pH conditions (acids, bases, alcohols), in our hands, the EOB fluorescence was too low for use in the quenching assay. Therefore, we employed an assay based on the characteristic shift in the absorbance maximum of EOB upon protein binding. The results showed that like FLN, EOB is a weak binder of Ab, with less than 3% of the dye bound even in the presence of an excess molar concentration of 25 mM Ab ( Figure S5; Panel B). The number of binding sites on Ab40 for the four dyes (RRB, EOY, ROB, and PHB) ranges between 1.5 and 2 suggesting that these dyes interact with multiple sites of Ab40. The multiple binding sites may explain different properties of the Ab aggregates induced by the dyes.
Since EOB and FLN displayed very poor binding to Ab and were also poor inhibitors of Ab aggregation, it clearly demonstrates that halogenation is very effective in generating molecules that tightly bind and consequently modulate the aggregation of Ab.

Heavy Halogen Atoms Play a Key Role in Modulating Ab Aggregation
Taken together, the TEM, CD, dot-blot, dye binding, and MTT reduction assay results indicate that FLN (negative control) without any halogen atom does not bind and modulate the Ab aggregation and cytotoxicity, whereas ERB congeners (ERB, EOY, PHB) containing multiple halogen atoms substantially modulated the Ab aggregation and effectively reduced the Ab cytotoxicity. Considering that FLN has a polyphenol-like structure but is a very poor Ab aggregation modulator, the molecular mechanism underlying the Ab aggregation modulation by ERB congeners was different from those of polyphenols. The assay results strongly support the idea that halogen atoms in the ERB congeners play an important role in the modulating Ab aggregation, and in the case of ERB, PHB, and EOY, ultimately Ab cytotoxicity. Having established this, the next issue becomes determining which specific features of halogen atoms are critical in modulating the Ab aggregation. From the CD, TEM, and dot-blot results of FLN and ERB analogs, several trends were found. First, the electronegativity of the halogen atoms/functional groups attached to xanthene group play an important role in Ab aggregation modulation. Although the results clearly show that EOY (which has four bromine atoms attached to the xanthene group) and ERB (which has four iodine atoms attached to the xanthene group) are both potent inhibitors of Ab fibril formation, ERB was slightly more effective than EOY at reducing the formation of fibrillar structures in the dot blotting and TEM assays. Furthermore, when the two bromine atoms close to benzoate group in the EOY structure are replaced with the two nitro groups in EOB, the inhibitory capacities of the small molecule on Ab fibril formation are eliminated. Therefore, the order of Ab fibril formation inhibitory capacity by xanthenes constituent group is I (ERB) .Br (EOY) .NO 2 (EOB). Because of this, either the electronegativity or size of the functional group attached to xanthene ring can be attributed to the inhibitory capacity of the ERB analogs. The order of the electronegativity and size of three atoms/groups is NO 2 .Br.I or NO 2 .I.Br, respectively. Therefore, we concluded that the inhibitory capacities are inversely proportional to the electronegativity of functional group attached to xanthene group rather than size, which is consistent with the recent findings on organofluorine Ab aggregation inhibitors [67]. Second, PHB and ROB (both of which contain four chlorine atoms on the benzoate group in addition to xanthenes group structures of EOY and ERB, respectively) led to the potent inhibitory capacities on Ab aggregation compared to the non-halogenated control molecules, EOB and FLN. This indicates that either polarity change or steric hindrance caused by four chlorine atoms added to the benzoate group resulted in the enhanced inhibitory capacities. However, ROB does not reduce Ab cytotoxicity, suggesting that both the location and type of halogen atoms on the xanthene benzoate affects the extent of Ab cytotoxicity inhibition.
Despite the two bromine atoms attached to xanthene benzoate group, EOB is not an effective modulator of Ab aggregation and cytotoxicity, Alternatively, we speculate that two nitro groups in EOB offset the positive effects of two bromine atoms on the modulatory capacity. Although more studies are required to clearly understand why EOB is not an effective modulator, other halogenated xanthene benzoates without any nitro group clearly exhibited the enhanced modulatory capacity on Ab aggregation over the xanthene benzoate without any halogen atom (FLN).

Conclusions
In this article, our investigation has conclusively established that ERB and two ERB analogs (EOY and PHB) effectively reduce Abassociated neurotoxicity by modulating Ab aggregation. In the case of ROB, while modulating capacities of ROB on Ab aggregation are prominent, it was not capable of alleviating Abassociated neurotoxicity. Comparative studies of ERB and ERB analogs on modulation of Ab aggregation and cytotoxicity revealed that FLN is not an effective modulator, but adding four heavy halogen atoms (either Br or I) to the xanthene group substantially enhanced the modulatory capacities on Ab aggregation and cytotoxicity. Adding four Cl atoms to the benzoate group also significantly enhanced the Ab aggregation modulation. In particular, co-incubation of PHB that contains four bromine atoms in the xanthene group and four chlorine atoms in the benzoate generates the low-molecular-weight Ab species with disordered structure similar to Ab monomer, which makes PHB a unique Ab aggregation modulator. Considering that halogen atoms play an important role in modulating Ab aggregation and cytotoxicity, ERB analogs are considered a new type of Ab modulators, halogenated small molecules. To our knowledge, this is the first report demonstrating the heavy halogen atoms added to multiple aromatic rings can confer inhibitory capacities on Abassociated cytotoxicity. Our studies can open a door to convert a poor Ab aggregation modulator into an effective one by adding heavy halogen atoms and serves as guidance to discover or design novel Ab aggregation modulators. Considering that ERB analogs are effective modulators of a-synuclein implicated in Parkinson's disease [68] and ERB itself is effective at destabilizing pre-formed Ab fibrils, [69], halogenation of small molecules might be a general way to obtain effective modulators of other amyloidogenic peptides and proteins at multiple stages of aggregation.

Supporting Information
Table S1 Spectral interference in the MTT absorbance by the residual dyes in the plate after washing. 1 st Row of Table S1 -Determination of the Dye Remaining in the Plate During the MTT Assay. The MTT assay was carried out as described previously in the MTT methods section, but with 10 mL of each dye-only control (3x concentration -no Ab) being added to each well. The absorbance of each dye was read at the respective absorbance maximum (ERB -540 nm, PHB -554 nm, EOB -520 nm, ROB -562 nm, EOY -530 nm, and FLN -492 nm) both before and after the washing steps described. After subtracting the appropriate background for both readings, the post-washing absorbance was normalized to the pre-wash absorbance in order to determine the fraction of each dye remaining after washing. 2 nd and 3 rd Rows of Table S1 -Determination of the Spectral Interference of the Dyes During the MTT Assay. To quantify the interference that varying fractions of residual dye remaining in the cell wells have on the final reduced form of MTT (MTTformazan) absorbance signal, fresh media was first added to a new cell culture plate without cells. Next, 7 mL of 1 mg/mL MTTformazan in DMSO was added to each well along with 0.01 and 0.05 fractions of each original dye amount or PBS. The absorbance of the samples was measured at 506 nm. After subtracting the background contribution of the media and DMSO, the absorbance values of the wells containing the varying dye fractions and MTT-formazan mixture were normalized to the wells with PBS/MTT-formazan to obtain the change induced in the MTT signal by the dyes left behind after washing (minimum triplicates tested). (DOC) Figure S1 MTT assay for ROB to Assess Viability of Neuroblastoma SH-SY5Y Cells. Three controls (PBS buffer, ROB, and Ab 0 d monomer) and two Ab aggregates formed in the absence (Ab 3 d) or presence (ROB Coincub) of 3x ROB at 37uC for 3 days. The Ab and ROB concentrations used were 5 and 15 mM, respectively (A). The Ab and ROB concentrations used were 2.5 and 7.5 mM, respectively (B). The ROB Bind sample refers to taking Ab 3 d aggregates formed in the absence of any dye and mixing them with 3x ROB immediately before addition to the cells. Values represent means 6 standard deviation (n$3).
Values are normalized to the viability of cells administered with PBS buffer only. Two-sided Student's t-tests were applied to the MTT reduction data. (Not significant: P.0.05). (TIF) Figure S2 Dot blot assay results using the A11 antibody. 50 mM of Ab monomer was incubated at 37uC in the absence (Ab only) or presence of 3x and 10x ERB analogs (EOB, EOY, ERB, ROB, and PHB) for up to 6 days. The samples were taken on the indicated day and the all samples were spotted onto one nitrocellulose membrane. The membrane was immuno-stained with the A11 antibody. For clearer presentation, the sections of the membrane were cut and re-arranged. (TIF) Figure S3 CD spectra of the Ab aggregates formed in the absence (Ab aggregates) or presence of 3x EOY, ERB, or ROB for 9 days at 376C. (TIF) Figure S4 Dot-blot assay results using the A11 antibody. The A11-reactive Ab aggregates (Ab at day 6), PBS buffer, and 10x ERB analogs were spotted into one nitrocellulose membrane.
Then, the membrane was immuno-stained with the A11 antibody. The sections from the same membrane were cut and re-arranged. (TIF)