Bifunctional fluorescent probes for detection of amyloid aggregates and reactive oxygen species

Protein aggregation into amyloid deposits and oxidative stress are key features of many neurodegenerative disorders including Parkinson's and Alzheimer's disease. We report here the creation of four highly sensitive bifunctional fluorescent probes, capable of H2O2 and/or amyloid aggregate detection. These bifunctional sensors use a benzothiazole core for amyloid localization and boronic ester oxidation to specifically detect H2O2. We characterized the optical properties of these probes using both bulk fluorescence measurements and single-aggregate fluorescence imaging, and quantify changes in their fluorescence properties upon addition of amyloid aggregates of α-synuclein and pathophysiological H2O2 concentrations. Our results indicate these new probes will be useful to detect and monitor neurodegenerative disease.


Introduction
Neurodegenerative diseases, such as Parkinson's and Alzheimer's disease, are characterized pathologically by the presence of amyloid deposits following protein misfolding, aggregation and neuronal loss in vulnerable brain regions [1]. Pathological aggregated proteins emerge by the misfolding of monomeric proteins into intermediate oligomeric structures, which polymerize into insoluble, highly organized structures, known as amyloid fibrils (figure 1a). There is a growing interest in the idea that these intermediate oligomeric species may play a crucial role in cellular toxicity in the disease [2]. Molecular alterations occur in the early stages of neurodegenerative disease and long precede symptomatic onset that accompanies neuronal death [3,4]. Thus, visualization of these early processes at the molecular level can be used to detect, predict and monitor the development of disease. In addition to protein aggregation, oxidative stress has been widely reported to occur in in vitro and in vivo models of neurodegenerative disease. Furthermore, oxidative damage is seen in post-mortem brain. Together, this evidence proposes a key role for oxidative stress in the pathogenesis of such diseases [5]. The interaction between α-synuclein (αSyn) and oxidative stress in neurodegeneration has been shown to be bidirectional, with oxidative stress promoting the aggregation of proteins such as αSyn and the misfolding of αSyn generates oligomers associated with increased reactive oxygen species (ROS) production [6,7]. Therefore, two crucial factors common in most neurological pathologies are (1) the accumulation of abnormally aggregated proteins and (2) high levels of ROS [8]. Abnormally aggregated proteins are commonly visualized and quantified using fluorescence-based methods. The main structural hallmark of amyloid-like aggregates is an ordered arrangement of β-sheets [9,10]; this structural feature facilitates specific binding of many fluorescent compounds including the widely used benzothiazole compound, thioflavin t (ThT) [11]. ThT undergoes an increase in fluorescence quantum yield (Φ Fl ) of several orders of magnitude upon incorporation into these β-sheet containing species [12]. This phenomenon occurs due to physical suppression of a non-fluorescent twisted intramolecular charge transfer (TICT) state as a consequence of rapid rotation about the C-C bond between benzothiazole and N,N-dimethylaniline rings (figure 1c) [12]. This fluorescence 'turn-on' makes ThT an essential probe of β-sheet in both the abundant fibrillar species and the rarer, potentially toxic oligomeric species [13,14].
ROS encompass a vast array of chemical compounds including singlet oxygen, hydroxyl radicals, peroxides and superoxides [15]; they are endogenously produced, reactive molecules and free radicals derived from molecular oxygen. In the brain they are principally generated by the electron transport chain in the mitochondria, the enzymes NADPH oxidase, xanthine oxidase and monoamine oxidases. Increased levels of ROS can chemically interact deleteriously with important biomolecules, causing protein carbonylation, DNA and RNA oxidation, DNA mutations [16] and lipid peroxidation, leading to organellar and cellular dysfunction and ultimately cell death. Oxidative stress in neurodegenerative diseases arises from an imbalance between the production of ROS and the level of antioxidants within the cell, leading to cell damage [8,17]. Previous methods to assess ROS levels are technically demanding [18] due to short half-lives and high reactivity, thus any sensor must function in a high spatial and temporal manner. Fortuitously, fluorescence imaging operates on a regime well suited to this [13] and is also often used in protein aggregation. Different classes of imaging probes such as nanomaterials and small molecule dyes have previously been used as fluorescence probes to study oxidative stress [19,20]. In this work, we demonstrate a new class of bifunctionalfluorescent small molecules that are capable of detecting both ROS and individual protein aggregates (figure 1b-d) A major and abundant ROS that is heavily involved in cell signalling processes is hydrogen peroxide (H 2 O 2 ). Additionally, the relatively high chemical stability of H 2 O 2 results in significantly increased in vivo, steady state, concentrations in neurodegenerative disorders [21]. Aryl boronic acids and esters have been shown to effectively and selectively react with H 2 O 2 and have thus been used broadly in drug delivery [22], as pro-chelators [23] and in H 2 O 2 imaging [24][25][26]. We have, therefore, designed a novel strategy for the concomitant detection of both ROS and amyloid aggregates (figure 1b). We designed four fluorescent probes: BE01, BE02, mBE01 and mBE02 (figure 1d), all composed of a benzothiazole core for amyloid targeting and an aryl boronic ester-based H 2 O 2 -sensitive unit (figure 1c). These four imaging probes were evaluated for their ability to function as a bifunctional fluorescence probe using both exogenous H 2 O 2 and aggregated recombinant αSyn protein, as a tractable model system.

Synthesis
Beginning from known bromide X [27], palladium-catalysed borylation provided either BE01 or BE02 in a single step in good yields (figure 1d). Thereafter, treatment with trimethyloxonium tetrafluoroborate (Meerwein's salt) provided the corresponding N-methylbenzothiazolium dyes mBE01 and mBE02. Each compound is an easily handled solid that can be prepared in useful quantities (greater than 100 mg) and the compounds were confirmed with NMR spectroscopy (electronic supplementary material, figures S1 and S2).

Bulk photophysical properties
To initially characterize the efficacy of these new probes, their bulk photophysical properties were assessed in physiologically relevant conditions (PBS, pH 7.4) prior to the addition of either preformed, recombinant αSyn aggregates [28] or H 2 O 2 . All probes showed absorption maxima in the ultraviolet region at 330 nm, with emission maxima between 412 and 417 nm. The cationic mBE dyes displayed no shift in absorption or emission wavelength relative to the neutral dyes (figure 2a-b) suggesting that the methylation does not alter the extent of delocalization in the chromophore. The fluorescence intensities of the free dyes in this aqueous buffer environment were weak, with all dyes having a Φ Fl < 0.1 (figure 2b). This observation was not wholly unexpected, due to both the electron withdrawing character of the boron and the non-emissive nature of the TICT state, like that observed with ThT [29,30]. The total integrated emission intensities (brightness) of the mBE dyes in solution were less than the neutral BE dyes, with higher ε values and lower Φ Fl , respectively. It is possible that the conformation in solution between the benzothiazole and phenyl-boronate rings in the methylated dyes remain somewhat, but not completely orthogonal (as in ThT) leading to low but detectable fluorescence at the bulk level [30].

Reactive oxygen species sensing
Next, the change in bulk fluorescence emission was assessed upon addition of H 2 O 2 at a concentration realistic for cellular oxidative stress (100 µM) [31]. We observed an immediate increase in bulk fluorescence from all dyes upon the addition of H 2 O 2 alone, (figure 2c) with mBE01 and mBE02 showing 16.0 and 3.6-fold increases and the neutral BE dyes showing more modest increases of 2.0-fold for BE01 and 1.3-fold for BE02 (figure 2d) . Furthermore, the kinetics could be followed, with all fluorescence traces increasing at varying rates in a linear regime over a 5 h period (figure 2c). As the oxidation reaction is predominantly irreversible, these sensors record the total ROS exposure to a system, rather than the equilibrium value. For BE01 we were able to detect H 2 O 2 to a concentration greater than 1 µM and simultaneously showed significant changes in fluorescence intensity between 1 and 100 µM H 2 O 2 (electronic supplementary material, figure S3). The oxidation of the electron withdrawing boronate ester to the electron donating hydroxyl was more significant with the cationic mBE dyes, hence the cationic dyes are more efficient sensors of H 2 O 2 alone. As expected, no change in ThT fluorescence intensity was observed in response to H 2 O 2 as no ester was present to be oxidized. To prove that the fluorescence changes seen are based on the oxidative cleavage of the boronic esters triggered by H 2 O 2 , identically prepared samples were analysed for their chemical composition via liquid chromatography mass spectrometry (LCMS, electronic supplementary material, table S1). These measurements confirmed that the change in fluorescence intensity arose from the conversion of the boronic ester into the hydroxyl functionality. By contrast, no significant change in the chemical composition of the dye solutions occurred over the same time frame without addition of H 2 O 2 (electronic supplementary material, figure S4). dyes showed affinity for these αSyn aggregates (figure 3), the mechanism of which is assumed to be analogous to ThT [32]. The capability of the dyes to bind αSyn aggregates was confirmed with surface plasmon resonance (SPR) measurements, in which αSyn aggregates were covalently attached onto the chip surface via an amide bond. The binding of the different dyes to the aggregates was then investigated taking ThT as a reference compound. Compared with ThT (K D = 51.2 ± 5.9 µM), all dyes showed improved binding to αSyn aggregates (figure 3). The non-methylated dyes (BE01 and BE02;

Aggregate sensing
3 µM) outperformed their methylated counterparts (mBE01 and mBE02; K D mBE01 = 35.2 ± 5.7 µM; K D BE02 = 33.5 ± 3.9 µM), which is in agreement with the literature comparing methylated and non-methylated benzothiazole compounds [33]. The K D of ThT has been reported to range from 0.033 to 64 µM [34]; we measured K D ThT = 51.3 ± 11.1 µM, towards the upper end of this range. This is probably due to the fact that SPR is a surface bound method where fixed, heterogeneous aggregates [13] are continuously exposed to a solvent flow, whereas fluorescence-based measurements, most commonly reported [34], are carried out in solution. These data indicate that any change in fluorescence was not due to affinity changes but rather a modification of a fluorescence property of the dye itself. Upon the addition of αSyn aggregates, we saw an increase in fluorescence intensity of 3.1, 4.8, 7.1 and 3.8-fold for BE01, BE02, mBE01 and mBE02, respectively (figure 2d, electronic supplementary material, figure S5).

Bifunctional sensing
All four dyes have demonstrated capabilities of independent fluorescence sensing of either H 2 O 2 or αSyn aggregates as the response to a chemical or physical modification of state. Excitingly, when H 2 O 2 was combined with αSyn aggregates we saw a large increase in the total fluorescence emission of the BE01 and BE02 compounds (figure 2d, electronic supplementary material, figure S5). In BE01 while addition of either H 2 O 2 or αSyn leads to a systematic increase in fluorescence of 2.0-fold and 3.3-fold, respectively, the addition of both together creates a 'compounded' increase of 7.5-fold. Similarly, in the BE02 case, H 2 O 2 by itself left the dye emission almost unchanged. However, an increase of 4.8-fold was observed upon the addition of αSyn aggregates, and the 'compounded' increase was significantly (electronic supplementary material, table S2 and figure S6) higher at 19.6-fold. As expected, ThT showed no change in fluorescence in the presence of coincident H 2 O 2 and αSyn aggregates. BE01 and BE02 showed significantly greater magnitudes of fluorescence intensity in comparison with mBE01 and mBE02, which showed no additional change in fluorescence in the bifunctional mode relative to αSyn aggregates alone. Taken together, these data support the idea that both BE01 and BE02 act as excellent bifunctional fluorescence sensors for concomitant detection of αSyn aggregates and H 2 O 2 at the bulk level. As the non-methylated BE dyes gave integrated intensity greater to that of ThT, we explored whether they were sensitive enough to function at the single-aggregate level.

Single-aggregate and H 2 O 2 fluorescence imaging
Next, the suitability of these probes was investigated for bifunctional in vitro studies at the singleaggregate level. We used a modified version of single-molecule total internal reflection fluorescence (TIRF) microscopy called single-aggregate visualization by enhancement (SAVE) imaging which we have previously demonstrated to detect and count single-aggregate species using ThT [13]. SAVE imaging was successfully applied to the new bifunctional probes, in all cases visualizing individual protein aggregates of αSyn on surfaces (figure 4a, electronic supplementary material, figure S7). This is exemplified in figure 4a, 15). Interactions of the BE and mBE dyes with αSyn showed fast exchange during the integration time of the detector, as with ThT (figure 4b) confirmed by constant, low fluorescence intensity over time and no apparent photobleaching [13]. Therefore, imaging for long periods of time would not be affected by the optical stability of these probes. To quantify this process, we determined the number of detectable aggregates, before and after addition of H 2 O 2 and assumed constant surface coverage. No αSyn aggregate species were visualized by BE01 under this imaging mode prior to oxidation; however, upon activation with H 2 O 2 , the density of species detected increased to 0.020 ± 0.004 aggregates/µm 2 ( figure 4c). BE02 showed a fivefold increase in the density of detectable aggregates from 0.005 ± 0.001 to 0.025 ± 0.006 aggregates/µm 2 . As a control, ThT displayed no change in species density (0.025 ± 0.004 and 0.026 ± 002 aggregates/µm 2 ). Interestingly, the densities of αSyn aggregates detected by oxidized BE01 and BE02 were similar to ThT despite the significantly higher fluorescence intensity of oxidized BE01 and BE02 with αSyn aggregates at the ensemble level. This suggests that both BE01 and BE02 are sensitive to similar species to ThT. There were very low densities of species detected by pre-oxidized mBE01 (0.0040 ± 0.0004 aggregates/µm 2 ) and mBE02 (0.006 ± 0.001 aggregates/µm 2 ) and negligible change in the density of species detected post-oxidation at 0.0051 ± 0.0004 aggregates/µm 2 and 0.0060 ± 0.0005 aggregates/µm 2 for mBE01 and mBE02 respectively. This agreed with the ensemble measurements and demonstrates the need for a neutral benzothiazole core for effective bifunctional sensing. The N-methylbenzothiazolium probes mBE01 and mBE02 can, however, be used in a limited capacity to measure either ROS activation or aggregate presence but do not show any advantage under exposure to both conditions. The limit of detection (LOD) of oxidized BE (BE-Ox) and mBE (mBE-Ox) probes to αSyn aggregates was determined using SAVE imaging (electronic supplementary material, figure S8). The LOD values are 4.58 ± 1.94 µM, 0.044 ± 0.005 nM and 2.40 ± 0.64 µM for BE-Ox, mBE-Ox and ThT, respectively, and were calculated by finding the first concentration at which the number of detected aggregates deviated significantly from the mean background (3 s.d.). The cellular biotoxicity of these probes was assessed for their potential use in vivo using a standard analytical cell-death assay (Trypan blue) and all probes both pre-and postoxidation were found to not affect cell viability after a 24 h period (electronic supplementary material, figure S9) [35].

Conclusion
New methods for the detection of ROS and protein aggregates will be critical to advance our clinical understanding of diseases. Accordingly, we have described the design and successful application of a new class of bifunctional fluorescent probes for the independent or simultaneous detection of both αSyn aggregates and H 2 O 2 . We used existing strategies for independently detecting either β-sheet containing aggregates of amyloid or ROS and united them to create four dyes BE01, BE02, mBE01 and mBE02. The BE and mBE dyes are distinguished only by N-methylation of the benzothiazole core, which provides the corresponding benzothiazolium fluorophores. The increased positive charge appears to have a profound effect on the bifunctional capabilities of the mBE probes relative to the neutral BE congeners. Ensemble measurements showed the mBE probes were competent independent H 2 O 2 or αSyn aggregate sensors but poor sensors of the two species simultaneously. This was further demonstrated with single-aggregate sensitivity. However, both BE01 and BE02 very successfully detected the simultaneous presence of both H 2 O 2 and αSyn aggregates, providing large increases in ensemble fluorescence intensity and detected single aggregates. These tools may prove useful in cellular studies of neurodegeneration where it is desirable to study both processes simultaneously.

Preparation of αSyn aggregates
Monomeric wild-type αSyn was purified from Escherichia coli as previously described in [36]. To generate amyloid fibrils a solution of αSyn (70 µM) in PBS buffer (pH 7.4) with 0.01% NaN 3 to prevent bacterial growth was incubated in the dark at 37°C with constant agitation (200 rpm) for several months before use. the dyes with αSyn aggregates. The dyes (25 µM) were then incubated with H 2 O 2 (100 µM) for 60 min. Following this incubation period the dyes were diluted to 5 µM and the emission spectra were recorded with αSyn aggregates. Spectra of the free dyes in PBS were also recorded at appropriate excitation wavelengths for calculation of Φ Fl relative to a standard [37].

Determination of ensemble photophysical properties
Several photophysical quantities were determined from ensemble absorption and emission data, using techniques previously described [38]. These included absorption maxima (λ Abs ), emission maxima (λ Em ) Stokes shift, molar extinction coefficient (ε) and fluorescence quantum yield (Φ Fl ). Stokes shifts were calculated from the difference of λ Em and λ Abs . ε was determined using the Beer-Lambert law with solutions of known concentrations. Φ Fl of the BE and mBE dyes in PBS were referenced against quinine sulfate (Sigma-Aldrich, Q0132) in 0.1 M H 2 SO 4 (Φ Fl = 0.5) [37], ThT was referenced against rhodamine 101 (Sigma-Aldrich, 83694) in ethanol (Φ Fl = 1) [39] and corrections were made for discrepancies in absorbance and solvent refractive index.

Surface plasmon resonance measurements
SPR studies were performed at 25°C using a Biacore T200 in HBS-EP + buffer (10 mM HEPES, 150 mM sodium chloride, 3 mM EDTA and 0.05% v/v surfactant P20 in ultra-pure, 18.2 MΩ cm water, Merck, SIMSV00WW) and filtered through a vacuum-driven filter system (0.22 µm, Merck) containing DMSO (5% v/v). αSyn aggregates, prepared as described previously, were covalently coupled to a research grade CM5 sensor chip (BIAcore) via primary amino functionalities of the fibrils using the amine coupling kit provided by BIAcore. For the coupling step, αSyn aggregates were injected at a concentration of 2.8 µM in sodium acetate buffer (10 mM, pH 4.0) at a flow rate of 10 µl min −1 for 12 min. Coupling levels ranged from 2146 to 11 350 RU. For reference subtraction, a second flow cell was treated identically including all immobilization chemistries but without αSyn aggregates.
At the beginning and at the end of each measurement set a solvent correction assay was performed to correct for the DMSO content in the samples. After stabilization of the baseline, response units for each dye sample were acquired by using a flow rate of 30 µl min −1 for 60 s. Dissociation time was set to 400 s. No further regeneration step was needed. Each dye concentration series was measured in a randomized order and the 10 µM sample was measured a second time after the whole concentration series of the dye to prove consistency. In between the measurements of different dyes a 5 µM ThT sample was measured as control sample. The whole SPR measurement set for each dye was repeated at least three times on different flow cells with fresh immobilized αSyn and new prepared concentration series. All data are solvent corrected and reference subtracted. The obtained binding levels from each concentration series were baseline corrected, normalized and the normalized values averaged over the replicates (n ≥ 3). Error bars represent twice the standard deviations of the binding level values between the replicates. The averaged values were then plotted as a function of concentration (figure 3). K D values were obtained by nonlinear curve fitting to these values using the equation Y = B max × X/(K D + X), where Y is the response, X the concentration, B max the maximum specific binding response (same unit as Y) and K D is the equilibrium binding constant (in the same units as X). The obtained fitting parameters are stated in electronic supplementary material, table S2.

Preparation for single-aggregate visualization by enhancement fluorescence imaging
Glass cover-slides for SAVE fluorescence imaging were cleaned for 1 h using an argon plasma (PDC-002, Harrick Plasma) and the slide surface incubated with poly-L-lysine (PLK) for 30 min as previously described [13]. Working solutions of dye (10 µM-1 pM) and αSyn (0.5 µM) in filtered (0.2 µm syringe filter, Whatman, 6780-1302) PBS (pH 7.4) were prepared 5 min prior to imaging. The working solution (50 µl) was added to the clean, PLK coated cover-slide and incubated for 2 min. The slide was then washed twice with PBS and imaged. H 2 O 2 (100 µM) was then added directly to the cover-slide and the sample imaged again.

SAVE fluorescence imaging
Fluorescence imaging experiments were carried out with a bespoke inverted microscope (Nikon, Eclipse, TE2000-U) set-up optimized for TIRF measurements. Excitation of BE, mBE compounds and ThT in the presence of αSyn aggregates were performed with a collimated 405 nm laser (Oxxius, Laserboxx, LBX-405-50-CIR-PP) which was aligned parallel to the optical axis and directed into an oil immersion objective lens (1.45 NA, 60×, Plan Apo, TIRF, Nikon) above the critical angle to ensure TIR at the coverslip-sample (glass/water) interface. Fluorescence emission was also collected by the same objective and selected by the presence of a dichroic (Di01-R405/488/561/635, Semrock) and subsequently passed through appropriate emission filters (BLP01-488R-25, Semrock). Images were then recorded by an EMCCD camera (Evolve delta 512, Photometrics) with an electron multiplication gain of 250 running in frame transfer, clear presequence mode. Each pixel on the image was 275 nm. Images were recorded at 100 ms exposure for 200 frames using automation through Micro-manager software [40].

Single-aggregate fluorescence image analysis
Images were analysed with a bespoke ImageJ [41] macro. Image stacks of 200 frames taken as 100 ms per frame of the BE, mBE dyes and ThT with αSyn aggregates and either in the presence or absence of H 2 O 2 were compressed in time to create single frame images representing the average pixel intensities. Pixels of intensity above background were determined with a fixed threshold and αSyn aggregate species in each field of view identified and counted. The density of species was determined from the division of the number of species by the area of the field of view (4726 µm 2 ). Five fields of view in the presence of H 2 O 2 and five in the absence of H 2 O 2 were analysed for each dye. Error bars represent the standard deviations of the species density between the five fields of view.

Cell-viability assay
HEK293 cells were maintained in high-glucose DMEM supplemented with 10% (vol/vol) FBS and 1% penicillin and streptomycin at 37°C and 5% CO 2 . Trypan blue exclusion was used to determine cell viability. Briefly, HEK293 cells were seeded at 80% confluence in six-well plates and treated with 10 µM BE01, BE02, mBE01, mBE02, BE-Ox and mBE-Ox separately. Cell viabilities were sequentially tested at 3, 12 (data not shown) and 24 h after the treatment. The total cell population and trypan blue stained cells were counted under a microscope. Healthy cell cultures were considered to be more than 85% viable cells as determined by the following equation: viable cells = 1 − number of blue cells total number of cells × 100 Data accessibility. This article has no additional data.