Few-layer bismuth selenides exfoliated by hemin inhibit amyloid-β1–42 fibril formation

Inhibiting amyloid-β (Aβ) fibril formation is the primary therapeutic strategy for Alzheimer’s disease. Several small molecules and nanomaterials have been used to inhibit Aβ fibril formation. However, insufficient inhibition efficiency or poor metabolization limits their further applications. Here, we used hemin to exfoliate few-layer Bi2Se3 in aqueous solution. Then we separated few-layer Bi2Se3 with different sizes and thicknesses by fractional centrifugation, and used them to attempt to inhibit Aβ1-42 aggregation. The results show that smaller and thinner few-layer Bi2Se3 had the highest inhibition efficiency. We further investigated the interaction between few-layer Bi2Se3 and Aβ1-42 monomers. The results indicate that the inhibition effect may be due to the high adsorption capacity of few-layer Bi2Se3 for Aβ1−42 monomers. Few-layer Bi2Se3 also decreased Aβ-mediated peroxidase-like activity and cytotoxicity according to in vitro neurotoxicity studies under physiological conditions. Therefore, our work shows the potential for applications of few-layer Bi2Se3 in the biomedical field.

a Wells-Dawson structure 26 have been used to inhibit Aβ fibril formation. However, there are also several challenges for nanomaterials to serve as Aβ inhibitors. The first problem is that these nanomaterials cannot be degraded, and are poorly metabolized. The second problem is that most of them fail to reduce Aβ -mediated neurotoxicity and peroxidase-like activity 27,28 . Furthermore, their inhibition mechanism has not been fully understood. To solve these problems, it is important to find new biocompatible materials that could be used to effectively inhibit Aβ aggregation.
Bismuth selenide (Bi 2 Se 3 ), a topological insulator, has attracted wide interest in condensed matter physics due to the unique surface electronic states [29][30][31] . It consists of stacked layers of a laminated structure held together by weak van der Waals interactions. The three-dimensional (3D) structure restricts its application due to the bulk state of high carrier density 32 . Thus, the production of two-dimensional (2D) Bi 2 Se 3 from its 3D bulk materials is in urgent demand in order to acquire superior property for potential applications. Up to now, 2D nanomaterials were mainly prepared by bottom-up synthesis and top-down exfoliation 33 . 3D materials with weak van der Waals forces can be exfoliated into thin flakes by mechanical or chemical exfoliation 34,35 , which is a top-down process. This method has been used to produce single-layer or few-layer 2D materials such as graphene and few-layer molybdenum sulfide because it is easier and more convenient than other methods 36,37 . Therefore, we try to prepare few-layer Bi 2 Se 3 by exfoliation of bulk Bi 2 Se 3 in solution. Though 2D few-layer Bi 2 Se 3 shows excellent properties which can be compared with graphene, there are only a few publications on the biomedical applications of Bi 2 Se 3 [38][39][40][41] . At the same time, the Se element can inhibit reactive oxygen species 42 . The Bi element, with atomic number 83, has a high photoelectric absorption coefficient and may be used as a cancer radio-sensitizer and X-ray contrast agent. Therefore, Bi 2 Se 3 has been reported to serve as a theranostic reagent for simultaneous cancer imaging and therapy 40 . More importantly, it has been reported that Bi 2 Se 3 nanoplates show low toxicity for mice even at high doses of 20 mg kg -1 . More surprisingly, Bi 2 Se 3 nanoplates can be metabolized after long-term toxicological responses 41 . These properties of 2D Bi 2 Se 3 stimulated us to investigate the interaction of 2D Bi 2 Se 3 and Aβ , and explore its ability to inhibit Aβ fibril formation.
In order to combine the two advantages of 2D nanomaterials and small molecules in inhibiting Aβ fibril formation, we prepared few-layer Bi 2 Se 3 coated by small molecules and investigated its ability to inhibit Aβ fibril formation. Here, few-layer Bi 2 Se 3 exfoliated by hemin ( Supplementary Fig. S1) was used to inhibit Aβ fibril formation under near-physiological conditions in vitro with satisfactory results. The preparation and application of exfoliated few-layer Bi 2 Se 3 are simple, low cost and environmental-friendly. These advantages indicate that our present work could provide new insights into the potential application of 2D few-layer Bi 2 Se 3 in medicine and biotechnology.

Results
Preparation and characterization of bulk Bi 2 Se 3 . Bulk Bi 2 Se 3 was prepared by hydrothermal synthesis. The as-synthesized bulk Bi 2 Se 3 exhibits sheet-like structure with a wide size distribution and is inclined to aggregate together, which was confirmed by scanning electron microscopy (SEM) and transmission electron microscopy (TEM) ( Supplementary Fig. S2a, b). The selected area electron diffraction (SAED) pattern ( Supplementary Fig. S2c) showed that Bi 2 Se 3 was indexed as a 6-fold symmetry [001] zone axis pattern, which is consistent with the layered structure along the z axis. Then, energy dispersive X-ray (EDX) spectrum was employed to confirm the elementary composition of Bi 2 Se 3 ( Supplementary  Fig. S2d). Furthermore, the thickness of bulk Bi 2 Se 3 was about 50 nm, as-determined by atomic force microscopy (AFM, Supplementary Fig. S2e). Finally, the as-synthesized bulk Bi 2 Se 3 was investigated by X-ray diffraction (XRD, Fig. 1f). All the labeled peaks were readily indexed as rhombohedral Bi 2 Se 3 (JCPDS no. 89-2008).
Optimization of exfoliation conditions. The as-prepared bulk Bi 2 Se 3 was firstly dispersed in hemin dissolved in 0.1% ammonia water (NH 3 ·H 2 O) solution. Then the mixture was sonicated for about 40 h to form few-layer Bi 2 Se 3 . Hemin, an iron-containing porphyrin small molecule, plays an important role in dispersing few-layer Bi 2 Se 3 . Therefore, it was necessary to investigate the effect of hemin concentration on the yield of few-layer Bi 2 Se 3 ( Supplementary Fig. S3). The results showed that the optimal concentration of hemin was 0.05 mg mL -1 . NH 3 ·H 2 O solution also plays an important role in dissolving hemin, which can affect the yield of few-layer Bi 2 Se 3 ( Supplementary Fig. S4a, b). 0.1% is the optimal concentration of NH 3 ·H 2 O. We further explored the effect of pH on the yield of few-layer Bi 2 Se 3 ( Supplementary Fig. S4c, d). The pH was adjusted by 0.1 M sodium hydroxide (NaOH) solution to substitute NH 3 ·H 2 O and 10 was the optimal pH that was close to the pH of 0.1% NH 3 ·H 2 O. We owe the exfoliation of bulk Bi 2 Se 3 materials to the energy provided by the ultrasound waves, which overcome the van der Waals forces between Bi 2 Se 3 layers. Therefore, the ultrasonic time is also an important parameter that should be investigated ( Supplementary Fig. S5). The yield of few-layer Bi 2 Se 3 was improved with increasing ultrasonic time; however, above an ultrasonic time of 40 h there was little change in the yield of few-layer Bi 2 Se 3 , and as such 40 h was chosen as the optimal ultrasonic time.
Characterization of few-layer Bi 2 Se 3 . Ultraviolet-visible (UV-vis) absorption spectra of the mixture of hemin and Bi 2 Se 3 before and after sonication were measured, respectively ( Supplementary Fig.  S6a). Before sonication, the solution exhibited a brown color and the spectrum of the mixture had a strong peak at 388 nm attributed to the Soret band of hemin, as well as a group of weak peaks between Scientific RepoRts | 5:10171 | DOi: 10.1038/srep10171 500 and 700 nm ascribed to the Q-bands of hemin 43 . After sonication of 40 h, the color of solution changed from brown to gray ( Supplementary Fig. S6b), maximum absorption band of the Soret band of hemin was red-shifted from 388 to 403 nm and a broad absorption in the visible light region appeared. These results are similar to few-layer Bi 2 Se 3 exfoliated by N-methyl-2-pyrrolidone in our previous work 44 , which could be attributed to the formation of few-layer Bi 2 Se 3 . At the same time, the absorption also increased gradually as the sonication time was extended ( Supplementary Fig. S5), which revealed that more few-layer Bi 2 Se 3 could be obtained with increasing sonication time. This suggests that the absorption was as a result of few-layer Bi 2 Se 3 .
The as-obtained few-layer Bi 2 Se 3 was a thin 2D flake according to the SEM (Fig. 1a) and TEM images (Fig. 1b). Furthermore, according to the SAED pattern ( Fig. 1b insert), few-layer Bi 2 Se 3 was indexed as a 6-fold symmetry [001] zone axis pattern, which is consistent with the layered structure along the z axis. Also, it revealed the single-crystalline nature of the thin 2D flake. The distance between the adjacently hexagonal lattice fringes investigated by the HRTEM was 0.207 nm for Bi 2 Se 3 (Fig. 1c), which is consistent with the lattice space of the (110) plane. The AFM image (Fig. 1d) also showed the flake structure and the thickness of exfoliated Bi 2 Se 3 was 3-4 nm (Fig. 1e), which nearly equals to 3-4 layers of Bi 2 Se 3 45 . The XRD pattern (Fig. 1f) of few-layer Bi 2 Se 3 showed a high degree of [001] orientation and some characteristic peaks disappeared compared to bulk Bi 2 Se 3 , which indicated that bulk Bi 2 Se 3 had been successfully exfoliated. To further confirm the exfoliation of Bi 2 Se 3 , Raman spectrum was employed ( Supplementary  Fig. S7). The A 1 1g mode of few-layer Bi 2 Se 3 produced a red shift compared with that of bulk Bi 2 Se 3 , which could be attributed to the phonon softening 46 . Furthermore, the content of hemin in few-layer Bi 2 Se 3 was 11.8%, as calculated by thermogravimetric (TGA) analysis (Fig. 1g), which is consistent with the calculated value (12.0%) by X-ray photoelectron spectroscopy (XPS) and EDX ( Supplementary Fig. S8).
Preparation of few-layer Bi 2 Se 3 with different thicknesses. Fractional centrifugation was employed here to obtain few-layer Bi 2 Se 3 with different layers. Firstly, the few-layer Bi 2 Se 3 was characterized by UV-vis absorption spectra ( Supplementary Fig. S9). Interestingly, we found that the dispersion solutions of few-layer Bi 2 Se 3 produced a broad absorption in the visible light region compared to bulk Bi 2 Se 3 . Furthermore, few-layer Bi 2 Se 3 stock solutions handled at different centrifugal speeds demonstrated different UV-vis absorption spectra. With centrifugal speed increasing from 2000 to 13000 rpm, the maximum absorption band of sample was blue-shifted gradually from 570 to 400 nm, resulting from quantum size effect 44 , which indicated the production of few-layer Bi 2 Se 3 with different sizes and layers. The size distribution and corresponding height profile of few-layer Bi 2 Se 3 collected at different centrifugation speeds were distinctive. TEM and SEM were used to measure the sizes of few-layer Bi 2 Se 3 (Fig. 2). With centrifugal speed increasing from 2000 to 13000 rpm, the size of sample decreased gradually from 637 ± 183 to 105 ± 31 nm, indicating the production of few-layer Bi 2 Se 3 with different sizes (Table 1). AFM is frequently used to character 2D materials. Here we used it to investigate the thickness of few-layer Bi 2 Se 3 . With centrifugal speed increasing from 2000 to 13000 rpm, the thickness of few-layer Bi 2 Se 3 decreased from about 40 to 3 nm (Fig 2.i-l), which further indicated the production of few-layer Bi 2 Se 3 with different layers. Inhibiting Aβ 1-42 fibril formation by few-layer Bi 2 Se 3 . Thioflavine T (ThT) is a classic amyloid dye that is frequently used to probe Aβ fibril formation due to its strong fluorescence emission upon binding to cross-β fibril structures [47][48][49] . We co-incubated Aβ 1-42 monomer and few-layer Bi 2 Se 3 with different concentrations, and then monitored Aβ fibril formation kinetics by ThT fluorescence assay. Modified Krebs-Henseliet buffer, which mimics near-physiological conditions 8 , was used in the following experiments except where specifically noted. Aβ 1-42 fibril formation in modified Krebs-Henseliet buffer without few-layer Bi 2 Se 3 was firstly investigated by ThT fluorescence assay ( Supplementary Fig. S10). In the absence of few-layer Bi 2 Se 3 , Aβ 1-42 formed ThT-positive β -sheets instantaneously and ThT fluorescence reached maximum intensity at 3 h and then decreased gradually. Therefore, 3 h was selected as the appropriate time to study the effect of few-layer Bi 2 Se 3 on Aβ 1-42 fibril formation. The Aβ fibril formation kinetics in the absence and presence of few-layer Bi 2 Se 3 with different thicknesses are shown in Fig. 3. Before fractional centrifugation, few-layer Bi 2 Se 3 with a wide thickness distribution (10 ± 8 nm) was named a mixture. When the mixture was introduced (Fig. 3a), the fluorescence intensity at 3 h gradually decreased with increasing concentration of few-layer Bi 2 Se 3 , indicating consistent inhibition of Aβ 1-42 fibril formation by few-layer Bi 2 Se 3 in a dose-dependent manner. To further investigate the effect of few-layer Bi 2 Se 3 thickness on Aβ 1-42 fibril formation, few-layer Bi 2 Se 3 with different thicknesses were introduced (Fig. 3b-d). Similarly, fluorescence intensities at 3 h gradually decreased with increasing concentration of few-layer Bi 2 Se 3 . It is interesting that the fluorescence intensity at 3 h gradually decreased with decreasing thickness of few-layer Bi 2 Se 3 at a same concentration, indicating the inhibition efficiency increased with decreasing layers of few-layer Bi 2 Se 3 ( Table 1 and Fig. 3f). Few-layer Bi 2 Se 3 contains 11.8% hemin calculated by TGA and XPS data ( Supplementary Fig. S8). In order to investigate the effect of hemin on Aβ 1-42 fibril formation, the corresponding hemin in few-layer Bi 2 Se 3 with different concentrations was calculated and incubated with Aβ 1-42 monomer in similar conditions ( Fig. 3e and Table 1). Compared with few-layer Bi 2 Se 3 , the decrease of ThT fluorescence intensity induced by hemin was negligible. Therefore, the high inhibition efficiency of few-layer Bi 2 Se 3 resulted mainly from few-layer Bi 2 Se 3 (Fig. 3f, Table 1). Finally, the end-point ThT intensities at 3 h versus different Aβ inhibitors were plotted (Fig. 3f) and few-layer Bi 2 Se 3 of 3 ± 1 nm had the best efficiency in inhibiting Aβ fibril formation. Therefore, few-layer Bi 2 Se 3 of 3 ± 1 nm was used in following experiments except where specifically noted.
To confirm that the reduced ThT fluorescence intensity resulted from the inhibiting Aβ 1-42 fibril formation by few-layer Bi 2 Se 3 , but not quenched by few-layer Bi 2 Se 3 and hemin themselves, we investigated the effect of few-layer Bi 2 Se 3 and hemin on ThT fluorescence intensity without Aβ 1-42 monomer ( Supplementary Fig. S11). The change of ThT intensity after incubation with few-layer Bi 2 Se 3 and hemin is unobvious. At the same time, the intrinsic fluorescence of hemin can also be neglected compared to ThT fluorescence of Aβ 1-42 fibril when the excitation wavelength is 442 nm (Supplementary Fig. S12). The results further suggest that the reduced ThT fluorescence intensity resulted from the inhibition of Aβ 1-42 fibril formation by few-layer Bi 2 Se 3 .
To further confirm the results of the ThT fluorescence assay, TEM and AFM were used to observe the morphologies of end-point products at 3 h (Fig. 4). After incubation of Aβ 1-42 monomer without few-layer Bi 2 Se 3 at 37°C for 3 h, Aβ 1-42 formed long, smooth, and entangled fibrils as expected (Fig. 4a, g). However, after incubation of Aβ 1-42 monomer and few-layer Bi 2 Se 3 with different concentrations, different Aβ species were observed. In the presence of 12 ng mL -1 few-layer Bi 2 Se 3 (Fig. 4b), fewer negatively-stained fibrils were observed and generally shorter in length than that without few-layer Bi 2 Se 3 . Aβ  in the presence of 60 ng mL -1 few-layer Bi 2 Se 3 formed some negatively-stained fibrils with amorphous aggregates (Fig. 4c, h) which were generally smaller in size than that without few-layer Bi 2 Se 3 . In the presence of 300 ng mL -1 few-layer Bi 2 Se 3 , primarily positively-stained aggregates with a small size were formed (Fig. 4d). However, Aβ 1-42 in the presence of 1200 ng mL -1 few-layer Bi 2 Se 3 formed small  particles adsorbed on few-layer Bi 2 Se 3 and no fibrils were observed (Fig. 4e, f, i). The SAED pattern ( Fig. 4f insert) indicated that the thin sheet was few-layer Bi 2 Se 3 . The AFM height profiles show that the size of Aβ 1-42 species decreased with increasing few-layer Bi 2 Se 3 concentration, which further demonstrates that few-layer Bi 2 Se 3 inhibits Aβ 1-42 fibril formation in a dose-dependent manner. The AFM height profile of Aβ 1-42 in the presence of 60 ng mL -1 few-layer Bi 2 Se 3 was higher than that of Aβ 1-42 without few-layer Bi 2 Se 3 , which could be attributed to adsorption of Aβ 1-42 aggregates on few-layer Bi 2 Se 3 surface. However, the AFM height profile of Aβ 1-42 in the presence of 1200 ng mL -1 few-layer Bi 2 Se 3 is close to that of bare few-layer Bi 2 Se 3 , which could be attributed to adsorption of Aβ 1-42 monomers on few-layer Bi 2 Se 3. TEM images of end-point products of Aβ 1-42 at 3 h in the presence of few-layer Bi 2 Se 3 mixture with a thickness 10 ± 8 nm also demonstrated inhibition of Aβ 1-42 fibril formation by few-layer Bi 2 Se 3 ( Supplementary Fig. S13). We employed dynamic light scattering (DLS) to study the size distribution of particles because DLS can provide a qualitative estimation of the aggregated state of Aβ 1-42 fibrils (Fig. 5). The freshly prepared Aβ 1-42 monomer in modified Krebs-Henseliet buffer has a hydrodynamic diameter around 4 nm (Fig. 5a). After incubation of Aβ 1-42 monomer without few-layer Bi 2 Se 3 at 37°C for 3 h, large particles with a diameter about 1000 nm were observed. When 12 ng mL -1 few-layer Bi 2 Se 3 was added, the peak at 1000 nm disappeared and broad peaks (400 nm -4 μ m) appeared, indicating parts of the Aβ 1-42 fibrils remained intact (Fig. 5c). At high concentration of few-layer Bi 2 Se 3 (1200 ng mL -1 ), no peak at 1000 nm was observed, indicating disappearance of large fibrils (Fig. 5d). The peak is almost the same as that of bare few-layer Bi 2 Se 3 (Fig. 5e). These results further indicate that few-layer Bi 2 Se 3 can inhibit Aβ 1-42 fibril formation. Furthermore, polyacrylamide gel electrophoresis (PAGE) and cyclic voltammograms (CVs) were employed to confirm the inhibiting effect of few-layer Bi 2 Se 3 on Aβ 1-42 fibril formation ( Supplementary  Fig. S14). Aβ 1-42 monomer, oligomer and fibril have different molecular weights. Freshly-prepared solution of Aβ 1-42 monomers displayed a strong band at 5 kDa, the molecular weight of Aβ 1-42 is 4.5 kDa. After incubation of Aβ 1-42 at 37°C for 3h, the monomer band (5 kDa) became weak and a band above 100 kDa appeared, indicating the formation of Aβ 1-42 fibrils. With the addition of few-layer Bi 2 Se 3 from 12 to 1200 ng mL -1 , the fibril band became weaker and weaker. When 1200 ng mL -1 few-layer Bi 2 Se 3 was added, the fibril band almost disappeared. To avoid the influence of SDS on Aβ 1-42 fibril formation, native PAGE was also performed ( Supplementary Fig. S14c). After incubation for 3 h, native PAGE indicated that Aβ 1-42 in the absence of few-layer Bi 2 Se 3 (lane 2) showed a decreased monomer band compared to freshly prepared Aβ 1-42 monomer (lane 1). However, when few-layer Bi 2 Se 3 with different concentrations were added, the monomer band recovered in a dose-dependent manner. The relative quantity of monomer band was calculated from Supplementary Fig. S14c. These results show that few-layer Bi 2 Se 3 can inhibit Aβ 1-42 fibril formation in a dose-dependent manner ( Supplementary Fig. S14d).
The conductivities of Aβ 1-42 monomer and fibril on the electrode surface should be different. Therefore, we also investigated the conductivities of glassy carbon electrode (GCE) modified by few-layer Bi 2 Se 3 , Aβ 1-42 monomer and end-products of Aβ 1-42 at 3 h with and without few-layer Bi 2 Se 3 ( Supplementary  Fig. S14b). GCE modified by few-layer Bi 2 Se 3 showed good conductivity almost the same as GCE. In the presence of Aβ 1-42 monomer, there was a small decrease in conductivity. However, a large drop for conductivity was seen when Aβ 1-42 fibril was introduced, which can be attributed to the insulating property of the fibrils. Interestingly, the conductivity recovered when Aβ 1-42 with few-layer Bi 2 Se 3 was introduced under the same concentration, further indicating that few-layer Bi 2 Se 3 might effectively inhibit Aβ 1-42 fibril formation, which might be due to the adsorption of Aβ 1-42 monomers on few-layer Bi 2 Se 3 to inhibit Aβ fibril formation.
Inhibition mechanism of Aβ 1-42 fibril formation by few-layer Bi 2 Se 3 . In order to investigate the inhibition mechanism, the inhibition process and the interaction of few-layer Bi 2 Se 3 and Aβ 1-42 were investigated by circular dichroism (CD), XPS, microbalance, CVs and electrochemical impedance (Fig. 6). In order to investigate whether modification of the secondary structure of Aβ 1-42 monomer occurred in the presence of few-layer Bi 2 Se 3 , CD spectra were collected (Fig. 6a). Freshly-prepared Aβ 1-42 monomer displayed a negative peak below 200 nm corresponding to random-coil structure of the peptide. After incubation of Aβ 1-42 at 37°C for 3 h, Aβ 1-42 in the absence of few-layer Bi 2 Se 3 displayed a negative peak at 217 nm corresponding to β -sheet structure of the peptide due to fibril formation. The peak was red-shifted to about 222 nm and a weak peak between 200 and 210 nm appeared with increasing concentration of few-layer Bi 2 Se 3 , which indicated the appearance of α -helix 50 . The percentage of different secondary structure for Aβ 1-42 monomer and Aβ 1-42 incubated at 37°C for 3 h in the presence and absence of few-layer Bi 2 Se 3 was calculated by Jasco secondary structure estimation software (Supplementary Table. S1). These results indicate that few-layer Bi 2 Se 3 may inhibit Aβ 1-42 fibril formation by preventing β -sheet structure formation.
In order to confirm that Aβ 1-42 monomer is adsorbed on the surface of few-layer Bi 2 Se 3 , CVs ( Supplementary Fig. S15) and Nyquist diagrams (Fig. 6b) of few-layer Bi 2 Se 3 -modified GCE before and after adsorbing Aβ 1-42 monomer were collected. Before adsorption, few-layer Bi 2 Se 3 -modified GCE showed good conductivity and the electron transfer resistance (R ct ) is 2700 Ω . After adsorption, the conductivity decreased and the R ct increased from 2700 to 7800 Ω , indicating that the Aβ 1-42 monomer had been successfully adsorbed on few-layer Bi 2 Se 3 . To further confirm the results, XPS and EDXA mapping images of few-layer Bi 2 Se 3 before and after adsorbing Aβ 1-42 monomer were collected. XPS data demonstrated that the carbon, nitrogen and oxygen contents of few-layer Bi 2 Se 3 after adsorbing Aβ 1-42 monomer increased from 48.87%, 5.9% and 21.72% to 54.90%, 10.3% and 30.25%, respectively ( Supplementary  Fig. S16), indicating adsorption of Aβ 1-42 monomer on the surface of few-layer Bi 2 Se 3 . EDXA mapping images showed that the Aβ 1-42 monomer was uniformly adsorbed on the surface of few-layer Bi 2 Se 3 ( Supplementary Fig. S17). These results motivated us to investigate the dynamics of the adsorbing process. First, few-layer Bi 2 Se 3 -modified GCE was prepared and immersed in modified Krebs-Henseliet buffer which buffer and Aβ 1-42 were added at different times, and impedance spectra with time was collected (Fig. 6c). No obvious impedance change was observed when the buffer was added at 30 and 90 min, while there was a large increase in impedance when Aβ 1-42 monomer was added at 60 min, which indicated that Aβ 1-42 monomer gradually adsorbed onto few-layer Bi 2 Se 3 . To quantify the adsorbing amount of Aβ 1-42 monomer, few-layer Bi 2 Se 3 -coated silicon wafer was hung on a microbalance and the weight was real-time monitored with Aβ 1-42 monomer adsorbed on the surface of few-layer Bi 2 Se 3 (Fig. 6d). The weight increased gradually with time and reached equilibrium after 20 min. In order to examine the mechanism and rate-controlling step in the overall adsorption process, pseudo-first-order and pseudo-second-order kinetic models were used to investigate the adsorption process. The nonlinear forms are expressed as the following equations, respectively 51 .
where ΔW 0 and ΔW t are the adsorption amounts of Aβ 1-42 monomer at equilibrium and at time t, respectively. k1 and k2 are the rate constants of pseudo-first-order and pseudo-second-order kinetic equation, respectively. Figure 6d shows the fitting curves by pseudo-first-order and pseudo-second-order kinetic equations. The fitting of pseudo-first-order kinetic curve overlaps with our experimental data and is better than that of pseudo-second-order kinetic curve. The kinetic parameters of two models are given in Table 2. The value of R 2 for equation (2) is 0.9343, and the calculated ΔW 0,cal is far from the experimental value of ΔW 0,exp . The experimental value is consistent with the calculated value from pseudo-first-order kinetic fitting, suggesting the adsorption process can be well-described by the pseudo-first-order kinetic model. The zeta potentials of Aβ 1-42 monomer and few-layer Bi 2 Se 3 are -16.3 and -29.4 mV, respectively ( Supplementary Fig. S18). Therefore, it is unlikely that the adsorption of Aβ 1-42 monomer is due to electrostatic interactions. These results suggest that the adsorption rate is mainly controlled by hydrophobic interactions between hemin on the surface of few-layer Bi 2 Se 3 and Aβ 1-42 .
To represent the suggested mechanism, a simple schematic for Aβ 1-42 fibril formation with and without few-layer Bi 2 Se 3 is depicted in Fig. 7. In the absence of few-layer Bi 2 Se 3 , Aβ 1-42 grows gradually into fibrils. In the 'nucleation phase' , Aβ monomers with random-coil structure undergo conformational change and aggregate into oligomers. In the 'elongation phase' , oligomers rapidly grow and form larger aggregates known as fibrils 52,53 . When few-layer Bi 2 Se 3 with a low concentration (<300 ng mL −1 ) is added, Aβ 1-42 monomers quickly adsorb on the surface of few-layer Bi 2 Se 3 and grow to form some aggregates due to the relatively high concentration of Aβ 1-42 , while free Aβ 1-42 monomers in solution aggregate into oligomers, then both Aβ 1-42 in solution and on few-layer Bi 2 Se 3 aggregate into fibrils. However, when few-layer Bi 2 Se 3 with a high concentration (>300 ng mL −1 ) is added, most of Aβ 1-42 monomers quickly  Table 2. Kinetic constants of the pseudo-first-order and pseudo-second-order kinetic models.
adsorb on the surface of few-layer Bi 2 Se 3 uniformly. Thus, the amount of free Aβ 1-42 monomer in solution is low and cannot aggregate to form fibrils.

Reducing Aβ-mediated peroxidase-like activity and cytoxicity. Recently, heme has been
reported to bind Aβ monomer, thus increasing peroxidase-like activity relative to free heme 28 . Therefore, we investigated the effect of few-layer Bi 2 Se 3 on Aβ -mediated peroxidase-like activity ( Supplementary  Fig. S19). Hemin-Aβ complexes showed remarkably enhanced peroxidase-like activity relative to free hemin, which is consistent with a study 54 . However, when free hemin is replaced by few-layer Bi 2 Se 3 containing the same amount of hemin, the enhanced peroxidase-like activity decreased to lower than that of free hemin, but a little higher than that of bare few-layer Bi 2 Se 3 . At the same time, few-layer Bi 2 Se 3 -Aβ complexes incubated at 37°C for 3 h also prevented enhanced peroxidase-like activity. Inhibition of enhanced peroxidase-like activity is attributed to the antioxidant effect of few-layer Bi 2 Se 3 . The results suggest that few-layer Bi 2 Se 3 could serve as an effective inhibitor of Aβ -mediated peroxidase-like activity.
To assess the cytoxicity effect of few-layer Bi 2 Se 3 -induced Aβ 1-42 species, we performed 3-[4,5-Dimethylthiazol-2-yl]-2,5-diphenyltetrazdium bromide (MTT) assays to examine the activity of mitochondrial alcohol dehydrogenase by treating rat glioma cells, C6, with our end-point products at 3 h. Cytoxicity was demonstrated by the reduction of cell viability and the viability is normalized to the one treated with the buffer control. Aβ 1-42 fibril alone contributed to ~22% cytoxicity to the glioma cells (Fig. 8a). For the end-point products obtained from Aβ 1-42 monomer incubated with few-layer Bi 2 Se 3 with different concentrations, the cytoxicity decreased with increasing concentration of few-layer Bi 2 Se 3  in a dose-dependent manner. When few-layer Bi 2 Se 3 concentration was 1200 ng mL -1 , the cytotoxicity was significantly reduced to ~7%. Furthermore, the toxicity of few-layer Bi 2 Se 3 itself was evaluated by MTT assay (Fig. 8b). Few-layer Bi 2 Se 3 itself with the concentration used above showed little cell toxicity under our experiment conditions. These results suggest that few-layer Bi 2 Se 3 can inhibit Aβ 1-42 -induced cell toxicity.

Discussion
In this paper, we successfully prepared few-layer Bi 2 Se 3 by liquid-exfoliation with the aid of hemin and found that it could exhibit good performance in inhibiting Aβ fibril formation. There are three reasons that hemin is selected as a stabilizer to exfoliate bulk Bi 2 Se 3 . First, hemin, iron protoporphyrin, is the active center of heme-proteins, such as cytochromes, peroxidase, myoglobin, and hemoglobin that are widely distributed in human body. Thus, hemin is a biocompatible molecule and suitable candidate as a stabilizer for few-layer Bi 2 Se 3 . It also contains a tetrapyrrole macrocycle, and the macrocycle is essentially planar structure, which trends to adsorb on the surface of 2D nanomaterials through π -π stack, hydrophobic and van der Waals interactions 55 . Therefore, hemin benefits the exfoliation of bulk Bi 2 Se 3 and stabilization of few-layer Bi 2 Se 3 . Second, the macrocycle of hemin also benefits the adsorption of Aβ 1-42 monomer through π -π stack, hydrophobic and van der Waals interactions because one Aβ 1-42 monomer contains four aromatic amino acids (three phenylalanines and one tyrosine) and three heterocyclic amino acids (three histidines) (Supplementary Fig. S20). Third, it has been reported that hemin inhibits Aβ aggregation 56 . Our experimental data in Fig. 3e also confirmed the result even though the inhibition efficiency was lower than that of few-layer Bi 2 Se 3 . Therefore, the synergistic effect of hemin and few-layer Bi 2 Se 3 might enhance the inhibition efficiency. The UV-vis spectra, XRD, SEM, TEM and AFM images of few-layer Bi 2 Se 3 indicated that few-layer Bi 2 Se 3 had been successfully prepared. XPS and TGA data show that 11.8% hemin was adsorbed on the surface of few-layer Bi 2 Se 3 . Furthermore, we used fractional centrifugation to obtain few-layer Bi 2 Se 3 with different sizes and thicknesses, which was confirmed by UV-Vis spectrum, SEM, TEM and AFM images.
Few-layer Bi 2 Se 3 with different thicknesses were used to evaluate their effect on inhibiting Aβ 1-42 fibril formation. ThT fluorescence assay was used to probe inhibition efficiency due to its strong fluorescence emission upon binding to cross-β fibrils [47][48][49] . The inhibition efficiency increased with increasing concentration of few-layer Bi 2 Se 3 . All few-layer Bi 2 Se 3 with different layers had higher inhibition efficiency than that of hemin, and the inhibition efficiency increased with decreasing layer of few-layer Bi 2 Se 3 , which could be attributed to the adsorption of Aβ monomer on the surface of few-layer Bi 2 Se 3 . The thin few-layer Bi 2 Se 3 had a larger specific area than that of thick few-layer Bi 2 Se 3 because specific area increases with decreasing layer of few-layer Bi 2 Se 3 , which benefits adsorption of Aβ 1-42 monomer. We propose that after adsorption on few-layer Bi 2 Se 3 , the concentration of free Aβ 1-42 monomer in solution is low and therefore there is a reduction in fibril formation. Aβ 1-42 monomers, oligomers and fibrils have different morphologies, sizes, molecular weight and conductivities. According to their different morphologies, TEM and AFM images (Fig. 4) support the result that few-layer Bi 2 Se 3 inhibits Aβ 1-42 fibril formation. According to their different sizes, DLS has showed that few-layer Bi 2 Se 3 inhibits Aβ 1-42 fibril formation (Fig. 5). According to their different molecular weights and conductivities, PAGE and CVs also confirm the inhibition effect of few-layer Bi 2 Se 3 on Aβ 1-42 fibril formation.
To understand the inhibition mechanism due to the adsorption of Aβ 1-42 monomer on few-layer Bi 2 Se 3 , we performed CD, CVs, electrochemical impedance, microbalance, XPS and EDX experiments. CD spectra showed that few-layer Bi 2 Se 3 prevented the β -sheet structure formation. To further understand this mechanism, CVs, Nyquist diagrams, XPS, EDX mapping before and after adsorption of Aβ 1-42 monomers were collected. The decreasing conductivity and increasing contents of carbon, nitrogen and oxygen support the result that Aβ 1-42 monomer was adsorbed onto the surface of few-layer Bi 2 Se 3 . EDX mappings also supported the result that Aβ 1-42 monomer was uniformly distributed on the surface of few-layer Bi 2 Se 3 . The dynamic adsorptive process was further examined by investigating the impedance change and weight increment with time. Adsorption reached equilibrium at 20 min, and the dynamic adsorptive process was well-described by a pseudo-first-order kinetic model. Furthermore, few-layer Bi 2 Se 3 reduced Aβ -mediated peroxidase-like activity and the resulting complex of Aβ 1-42 and few-layer Bi 2 Se 3 reduced glial C6 cells toxicities in comparison to those treated with mature Aβ 1-42 fibrils formed in the absence of few-layer Bi 2 Se 3 . The results suggest that the negatively charged few-layer Bi 2 Se 3 in the microgram range could potentially serve as an Aβ inhibitor to prevent cell death caused by those detrimental Aβ species.
In summary, we have prepared a new Aβ inhibitor by a simple process. The most important discovery is that few-layer Bi 2 Se 3 might serve as an Aβ inhibitor and exhibit higher inhibition efficiency against Aβ 1-42 fibril formation than hemin. The high inhibition efficiency of few-layer Bi 2 Se 3 may be attributed to its high adsorption capacity for Aβ 1-42 monomer. Furthermore, few-layer Bi 2 Se 3 , with a good biocompatibility, may reduce enhanced peroxidase-like activity and the cytoxicity induced by Aβ 1-42 . We can envision that few-layer Bi 2 Se 3 may have the potential for applications in the biomedical field. The metabolism of few-layer Bi 2 Se 3 in vivo and its possible biomedical applications will be investigated in our future work.
Scientific RepoRts | 5:10171 | DOi: 10.1038/srep10171 Methods Synthesis of bulk Bi 2 Se 3 . Bulk Bi 2 Se 3 was synthesized as we previously reported 44 . Polyvinyl pyrrolidone (0.9 g) was dissolved in 36 mL of ethylene glycol (EG). Then bismuth oxide powder (1 mmol), selenium powder (3 mmol) and ethylenediamine tetraacetic acid powder (4 mmol) were added into above-mentioned EG. The resulting suspension was stirred vigorously and subsequently sealed in a steel autoclave. The autoclave was heated to 200°C in 30 min and maintained over 20 h for a completed reaction. The as-obtained product was collected by centrifugation at 8000 rpm for 15 min, then washed three times with deionized water and absolute ethanol, and finally dried at 60°C for 96 h in an oven.
Preparation of few-layer Bi 2 Se 3 . Stock solution of hemin (0.05 mg mL -1 ) was prepared in 0.1% NH 3 ·H 2 O aqueous solution. The prepared bulk Bi 2 Se 3 (100 mg) was dispersed in 200 mL of hemin solution by sonication for 40 h in a sonic bath (KQ-250 DB, 250 W). The resulting dispersion was left to stand for 48 h to allow any unstable aggregates to form. The supernatant was centrifuged at 2000 rpm for 20 min. The precipitate was collected as sample 2000 rpm. After centrifugation, the supernatant was collected and centrifuged at 6000 rpm for 20 min, and the precipitate was collected as sample 6000 rpm. Then, the remaining supernatant was centrifuged at 10000 rpm for 20 min, and the precipitate was collected as sample 10000 rpm. Finally, the collected supernatant was further centrifuged at 13000 rpm for 20 min, and the precipitate was collected as sample 13000 rpm. Effect of few-layer Bi 2 Se 3 on Aβ 1−42 fibril formation. The previously prepared Aβ 1-42 monomer solution was mixed with designed volumes of few-layer Bi 2 Se 3 (6 μ g mL -1 ) solution and then diluted by sterilized modified Krebs-Henseliet buffer to the final solution of 10 μ M Aβ 1-42 with 0, 12, 60, 300 and 1200 ng mL -1 few-layer Bi 2 Se 3 , respectively. The samples were incubated at 37°C in 1.5 mL of Eppendorf tubes from 0 to 3 hours, and some of samples were taken out at the desired incubation time and used for the following measurements of ThT fluorescence, TEM, AFM, DLS, SDS-PAGE, CD, CVs, Nyquist diagrams and MTT assay. The physiologically-significant medium was a modified Krebs-Henseliet buffer including 118.5 mM sodium chloride, 4.8 mM potassium chloride, 1.2 mM magnesium sulfate, 1.4 mM calcium chloride, 11.0 mM glucose and it was buffered at pH 7.40 ± 0.05 with 100 mM Piperazine-1,4-bis(2-ethanesulfonic acid) 58 . The buffer also included 0.05% w/v sodium azide as an antimicrobial. All treatments were prepared and maintained at 37°C to accurately reflect physiological milieu.

Characterization of bulk Bi
ThT fluorescence assay. The fibril formation of Aβ 1-42 with or without inhibitors was evaluated by ThT fluorescence with a fluorescence spectrophotometer (FLS-920, Edinburgh Instruments). 50 μ L of Aβ 1−42 (200 μ M) was mixed with few-layer Bi 2 Se 3 with different layers of required concentration and incubated in a 1.5 mL of sterile brown centrifuge tube, which included 10 μ M ThT and 0.05% sodium azide in the modified Krebs-Henseliet buffer. The final volume was 1 mL. The Aβ concentration was fixed at 10 μ M (45.14 μ g mL -1 ) and concentration of few-layer Bi 2 Se 3 was changed from 0 to 1200 ng mL -1 . The samples were loaded on a thermomixer (Eppendorf, Germany) at 37°C for 0-3 h during which samples were collected and measured. Three replicates were performed and averaged.
TEM and AFM measurements. Final solution of 10 μ M Aβ 1-42 with 0, 12, 60, 300 and 1200 ng mL -1 few-layer Bi 2 Se 3 incubated at 37°C for 3 h were used for TEM and AFM measurements. The morphologies of Aβ 1−42 with and without few-layer Bi 2 Se 3 were confirmed by TEM and AFM. For TEM measurements, samples were deposited on 400-mesh Formvar carbon-coated copper grids for 5 min. Negative staining was performed by using 2% uranyl acetate for 5 min and the grids were rinsed once with double distilled water. The samples were examined with a JEOL JEM-2100 TEM with an accelerating voltage of 200 kV. For AFM measurements, 10 μ L of samples were freshly prepared and swiftly diluted 10-fold in deionized water. 10 μ L of the diluted sample was mounted onto the freshly cleaved mica for 5 min, gently rinsed with deionized water, and dried in vacuum overnight. Images were acquired under atmosphere in a tapping mode.
SDS-PAGE and Native-PAGE analysis. 20 μ M Aβ 1-42 monomer, 20 μ M Aβ 1-42 incubated at 3 h with 0, 24, 120, 600 and 2400 ng mL -1 few-layer Bi 2 Se 3 were used for SDS or Native-PAGE analysis. For SDS-PAGE analysis, the samples (10 μ L) were mixed with 2× PAGE sample loading buffer (10 μ L). Then samples were run on 12% PAGE gel at 80 V for 0.5 h followed by 120 V for 1.5 h. The gel was stained by coomassie blue. For Native-PAGE analysis, samples were run on 12% Tris/glycine gel at 100 V for 10 min followed by 200 V for 0.5 h. The gel was stained using silver stain and the relative quantity was estimated using Bio-Rad's Image Lab 4.1 software. The bands were both visualized using gel imaging and analysis system (Bio-rad Gel Doc XR). CD spectra. 20 μ M Aβ 1-42 monomer, 20 μ M Aβ 1-42 incubated at 3 h with and without 0, 120, and 2400 ng mL -1 few-layer Bi 2 Se 3 were used for CD spectra. The CD spectra were measured from 190 to 260 nm at room temperature on a Jasco J-810 spectrometer (Tokyo, Japan) using a cell with a path length of 0.1 cm. Data were collected every 0.2 nm with 3 nm bandwidth at a scan speed of 50 nm min -1 and response time of 4 seconds. All spectra were collected in a triplicate and a background-corrected against buffer blank.
Inhibiting Aβ 1-42 fibril formation by CVs. 10 μ M of Aβ 1-42 monomer, 10 μ M Aβ 1-42 incubated at 37°C for 3 h with 0, and 1200 ng mL -1 few-layer Bi 2 Se 3 , and 1200 ng mL -1 few-layer Bi 2 Se 3 were used for CVs measurements. The glass-carbon electrode (GCE, Φ = 3 mm) was polished with 0.3 and 0.05 μ m alumina slurry, rinsed thoroughly with doubly distilled water between each polishing step, then GCE was washed successively with 1:1 nitric acid, ethanol, and doubly distilled water in an ultrasonic bath and dried in air. 10 μ L of samples were dropped onto the GCE to prepare modified electrodes. CVs measurements were performed on an electrochemical workstation (CHI660C, CH Instrument, USA). The three-electrode system consisted of a platinum wire as auxiliary electrode and an Ag/AgCl (saturated KCl) as reference. Working electrodes were GCEs modified with samples. CV measurements were performed in 6.0 mM K 3  Adsorption by CVs and impedance measurements. 10 μ L of few-layer Bi 2 Se 3 (10 μ g mL -1 ) was dropped onto the GCE to prepare few-layer Bi 2 Se 3 -modified electrode. For adsorbing Aβ 1-42 monomer, few-layer Bi 2 Se 3 -modified electrode was immersed in modified Krebs-Henseliet buffer containing 10 μ M Aβ 1-42 monomer for 10 min, and then immersed in modified Krebs-Henseliet buffer for 30 seconds to remove free Aβ 1-42 monomers. Few-layer Bi 2 Se 3 -modified electrode before and after adsorbing Aβ 1-42 monomer were used for CVs and impedance measurements. CVs were performed almost the same as above. The impedance measurements were performed in 5.0 mM [Fe(CN) 6 ] 3-/4and 1.0 M KCl solution. The AC voltage amplitude was 5 mV and the voltage frequencies ranged from 0.1 Hz to 10 5 Hz.
Adsorption kinetics process by impedance measurements and real-time weight measurements. The impedance measurement with time was performed in modified Krebs-Henseliet buffer at 10 Hz. Few-layer Bi 2 Se 3 -modified electrode was subjected to successive incubations first in modified Krebs-Henseliet buffer without Aβ 1-42 as a pre-equilibration step. Buffer was added twice at 30 min and 90 min, respectively. Aβ 1-42 monomer was added at 60 min.
Real-time weight measurement was performed on a microbalance (KSV NIMA). Silicon wafer was washed successively with 1:1 nitric acid, ethanol, and doubly distilled water in an ultrasonic bath and dried in air. 50 μ L of few-layer Bi 2 Se 3 (10 μ g mL -1 ) dispersion was dropped onto silicon wafer to prepare few-layer Bi 2 Se 3 -coated silicon wafer. Then the modified silicon wafer was dried naturally overnight. Then few-layer Bi 2 Se 3 -coated silicon wafer was hung on a microbalance and immersed in modified Krebs-Henseliet buffer containing 10 μ M Aβ 1-42 monomers. Thus, the weight of adsorbed Aβ 1-42 monomers was in situ monitored.
Inhibition of Aβ-mediated peroxidase-like activity. The inhibition assay of Aβ -mediated peroxidase-like activity was performed as follows. Firstly, 12 μ g mL -1 of few-layer Bi 2 Se 3 or 1.416 μ g mL -1 of hemin in the presence and absence of 50 μ M Aβ 1-42 monomers were incubated at 37°C for 0 h or 3 h in a 1.5 mL of sterile centrifuge tube and served as catalyst for peroxidase-like activity assay. Kinetic measurements were carried out in time course mode by monitoring the absorbance change at 652 nm on a TU-1901 UV-vis spectrophotometer. Peroxidase-like activity experiments were performed using the above samples as catalyst in a reaction volume of 600 μ L acetic acid-sodium acetate buffer solution (0.02 M, pH 4, 25°C) with 100 μ M 3,3′ ,5,5′ -Tetramethylbenzidine as substrate and 10 mM hydrogen peroxide.

Inhibition of Aβ-mediated cytotoxicity. The rabbit glial C6 cells (from cell storeroom of Chinese
Academy of Sciences) were incubated at 37°C under 5% CO 2 , and were cultured in DMEM media (Gibco, Invitrogen, Carlsbad, CA, USA) with 10% fetal bovine serum (FBS) (Biological Industry, Kibbutz Beit-Haemek, Israel) in a humid camber. A total of 1× 10 5 cells were seeded overnight in the growth medium in a polystyrene 96-well plate (Corning, NY, USA). The growth medium was then discarded and the cells were washed twice by 1× phosphate buffered saline (PBS). 1× PBS was prepared by dissolving 0.24 g KH 2 PO 4 , 1.44 g Na 2 HPO 4 , 0.2 g KCl and 8.0 g NaCl into 1 L of water. pH was adjusted to 7.4 with HCl. The final concentration was 10 mM phosphate, 137 mM NaCl and 2.7 mM KCl. Then, the FBS-free media (50 μ L) was added into each well. Next, the cells were treated with 50 μ L of the end-point products as described in text. The cells were incubated for additional 24 h in the growth chamber and then 50 μ L of MTT (5 mg mL -1 in DMEM without FBS) was added into each well and incubated for another 4 h. The media was discarded and dimethylsulfoxide was used to lyze the cells until the purple crystals were fully dissolved. Absorbance at 570 nm was measured by a microplate reader (SpectraMax M5, Molecule Devices). Three replicates were performed and the data were averaged (n = 3). Background signals from sample treatment without cells were subtracted. Each data set was normalized using the reading obtained from the buffer controls and the cytotoxicity data were obtained by subtracting the viability data from 100%.