Probing supramolecular protein assembly using covalently attached fluorescent molecular rotors

Changes in microscopic viscosity and macromolecular crowding accompany the transition of proteins from their monomeric forms into highly organised fibrillar states. Previously, we have demonstrated that viscosity sensitive fluorophores termed 'molecular rotors', when freely mixed with monomers of interest, are able to report on changes in microrheology accompanying amyloid formation, and measured an increase in rigidity of approximately three orders of magnitude during aggregation of lysozyme and insulin. Here we extend this strategy by covalently attaching molecular rotors to several proteins capable of assembly into fibrils, namely lysozyme, fibrinogen and amyloid-β peptide (Aβ(1-42)). We demonstrate that upon covalent attachment the molecular rotors can successfully probe supramolecular assembly in vitro. Importantly, our new strategy has wider applications in cellulo and in vivo, since covalently attached molecular rotors can be successfully delivered in situ and will colocalise with the aggregating protein, for example inside live cells. This important advantage allowed us to follow the microscopic viscosity changes accompanying blood clotting and during Aβ(1-42) aggregation in live SH-SY5Y cells. Our results demonstrate that covalently attached molecular rotors are a widely applicable tool to study supramolecular protein assembly and can reveal microrheological features of aggregating protein systems both in vitro and in cellulo not observable through classical fluorescent probes operating in light switch mode.


Table of Contents page
1) Additional spectroscopic and imaging data 2      We have recorded time resolved decays of sulfo-Cy3-HEWL in various sucrose/water solutions at variable temperatures and plotted them in the phasor space, Figure 2b (main text). In addition we have also performed exponential analysis of the individual decays and calculated the mean fluorescence lifetime, Figure S6. While the data overlap well in the phasor space ( Figure 2b, main text), exponential fitting revealed a dependence of the position of the mean fluorescence lifetime on both the temperature and the viscosity of the mixture ( Figure S6). Such behaviour is not unusual and was reported previously for molecular rotors. 1 Thus, for calibration purposes, in order to remove the bias associated with temperature variations in the calibration set, we have recorded time resolved fluorescence decays of sulfo-Cy3-HEWL in sucrose/water mixtures at a fixed temperature of 60°C, the intended temperature of the aggregation reaction. The time resolved decays are shown in Figure 2a (main text) and the mean fluorescence lifetime calibration of the data is plotted in Figure S6 (black squares). From this dataset it is clear to see that an increase in viscosity causes a significant increase in the fluorescence lifetime of sulfo-Cy3-HEWL. Figure S7. Phasor analysis of HEWL aggregation monitored by freely added DiSC2(3) (blue) compared to sulfo-Cy3-HEWL (red). Only the centroids of phasor clouds are shown for clarity; the arrow shows the direction of the trend. The dynamic range of the conjugated dye is smaller than in the case of free dye. Figure S8. The disruption of blood clots after the application of tissue plasminogen activator (TPA). Confocal fluorescence images of fibrinogen clots obtained at time 0, 1/2 h and 1h following the addition of the known clot-disrupting drug TPA (1.5 μg/ml) to the incubating solution. Excitation was at 1000 nm, detection was 600-700 nm. It is clear that the clot starts dissolving following the addition of the drug as monitored by the sulfo-Cy3 fluorescence. The fact that sulfo-Cy3 was covalently linked to fibrinogen was confirmed by the formation of a fluorescent fibrin mesh after blood clotting (a). Further evidence is provided by an experiment in which the fluorescent mesh containing Cy3-fibrin was disassembled due to the application a clot-dissolving drug, TPA. The effect can be clearly seen after ½-1 hour of the application of TPA (b and c, respectively). Figure S9. Phasors corresponding to decays of sulfo-Cy3-fibrin conjugates in human blood clots compared to porcine plasma clots. In both cases, the phasor point recording after clotting lies anticlockwise from the phasor recorded before clotting, indicating a higher degree of crowding as expected. Lysozyme aggregation is monitored by BODIPY-HEWL (red) and sulfo-Cy3-HEWL (blue), Aβ(1-42) aggregation by BODIPY-Aβ(1-42) (green) and fibrin clotting by sulfo-Cy3-fibrinogen (orange). Clearly, the fluorescence lifetime sensitivity to microviscosity is very dependent on the specific rotor-protein system. Sulfo-Cy3-HEWL has the biggest dynamic range during the assembly of HEWL into fibrils, but the same rotor (sulfo-Cy3) in the sulfo-Cy3-fibrinogen system during clotting has the smallest dynamic range. This may be due to Cy3 rotation being largely hindered from the start, upon covalent attachment (as indicated by its initial high fluorescence lifetime); however, we believe that the rotor may not be ideally located on a suitable residue(s) on the protein, in order to strongly sense the crowding associated with fibrin mesh assembly.
All phasors are located within the universal circle, indicating that the rotors are experiencing an inhomogeneous environment, from the very start (i.e. conjugation) through to fibril formation. While the trends for sulfo-Cy3-HEWL (blue), BODIPY-Aβ(1-42) (green) and sulfo-Cy3-fibrinogen (orange) appear to proceed in a linear fashion in the phasor space (i.e. possibly involve the conversion between two species), the trend for BODIPY-HEWL (red) is clearly biphasic, see Figure S12 for more details.

Monitoring lysozyme aggregation using BODIPY-HEWL
We have tested whether it is possible to monitor protein aggregation with a hydrophobic BODIPY-based molecular rotor, previously used for a range of biological viscosity studies in lipid-based membranes. 2-10 BODIPY-C10 ( Figure S1c, Supplementary Material) displays a large dynamic range of lifetimes in the viscosity range between 20-5000 cP and is characterised by monoexponential time resolved fluorescence decays in homogeneous media, 8,9 and was shown to measure viscosity in temperature-independent manner. 1 However, the use of BODIPY-based rotors in water-based systems was precluded due to a poor solubility of the dye in an aqueous media.
We have synthesised BODIPY-NHS (Figure 1b   The viscosity sensitivity of the BODIPY-HEWL conjugates was confirmed by measuring the time resolved fluorescence decays in water/glycerol mixtures of varied viscosities ( Figure S12a). It is clear to see that increasing viscosity resulted in increasing fluorescence lifetime of BODIPY-HEWL, indicating that the viscosity sensitivity of BODIPY was retained even after conjugation.
Lysozyme aggregation was induced in the same manner as for Cy3-HEWL and fluorescence decays traces of BODIPY-HEWL were recorded during the course of aggregation, until fibril formation was apparent and there was no further change in the time resolved traces recorded from the mixture. As expected, the mean fluorescence lifetime increased throughout aggregation, proving that BODIPY conjugates are effective for monitoring the aggregation process ( Figure 5b, main text).
The time resolved decays recorded for free BODIPY-C10 in methanol/glycerol mixtures of different viscosities can be fitted with a monoexponential function 7,8 and therefore fall onto the universal circle in phasor space (Figure 11b, grey). By contrast, the BODIPY-HEWL conjugate shows a substantial deviation from monoexponentiality and, in addition, the dynamic range of lifetimes is smaller for the conjugate than for the free dye, likely due to an increased rotation barrier caused by conjugation. While the more complex shape of decays can cause difficulties in analysis, it does not preclude the monitoring of the aggregation using BODIPY-HEWL.
To ascertain that the high fluorescence lifetime observed for BODIPY-HEWL during aggregation is not due to an artefact we have measured the time resolved fluorescence decay of BODIPY-C10 in the glassy matrix of ethanol-methanol at 77K with a known dynamic viscosity of η=2x1014 cP, where we expect the rotor to be completely immobile. This decay can be fitted with a monoexponential function and is characterised by a lifetime of 6 ns, 11 which lies anticlockwise from the extrapolated point C' of BODIPY-HEWL aggregation. Thus we confirm that BODIPY is not completely immobilised in the final fibrillar state of HEWL, since it displays lower lifetime than in a solid matrix at 77K and thus it is in principle capable of sensing even higher microscopic viscosity than that achieved during the aggregation of HEWL. Two stages can be distinguished in the evolution of phasors during aggregation, which become apparent from the direct analysis of phasor clouds ( Figure S13). In the first minutes after seeding the phasor clouds proceed along a line (A to B), which can be interpreted as a transition between two states of different microviscosity (or crowding) experienced by BODIPY (with viscosities characterised by points A' and B' on the universal circle). 12 After approximately 6 minutes, the trend changes and the phasor clouds start proceeding along the second line (B to C), which again in the phasor space indicates an interconversion between two species.
The transition towards a third state characterised by the viscosity value corresponding to C' can be explained by the appearance of a new species of aggregate. If extrapolated towards the universal circle (C'), it is apparent that the trend is evolving towards a region of very high fluorescence lifetime, which, nevertheless, is less than the maximum achievable lifetime of BODIPY rotor recorded in a solid glass matrix at 77K, 6 ns (point D, Figure S13 c).  For the first 2h both samples (a) and (b) are mainly comprised of oligomeric species, although a very small number of short fibrils are already present at the 1 hr time point.
The first Aβ(1-42) fibrils seem to appear slightly later than the labelled BODIPY-Aβ(1-42) fibrils. However, this is likely due to the fact that BODIPY-Aβ(1-42), has already started aggregating during a conjugation reaction, which was allowed to run for 1 hour.
After ca 4 hours the formation of first long fibrils (>1 µm) can be clearly seen in both unlabelled and labelled Aβ(1-42) samples.   In the case of sulfo-Cy3-NHS conjugation to HEWL Fluorescence lifetime imaging (FLIM) was used to monitor the aggregation process as described below (in 'Time resolved measurements' section).

Protein aggregation
A fresh solution of HEWL was prepared at 4 mg/mL with 50 mM HCl, 100 mM NaCl. The monomer solution was passed through a syringe driven filter of 0.22 μm pore size. A fraction of sulfo-Cy3-HEWL conjugates prepared as described above was added to the fresh solution of unlabelled proteins. Aggregation was triggered by seeding with 1 % v/v sonicated preformed fibrils. The seeds were prepared by incubating a solution of 60 mg/mL HEWL in 50 mM HCl, 200 mM NaCl at 65°C for 16 hours and subsequent sonication to obtain short seed fibrils. The pH of the mixture was less than 2 for all experiments. Fluorescence lifetime imaging (FLIM) was used to monitor the aggregation process as described below (in 'Time resolved measurements' section).

Viscosity calibration of rotor-HEWL conjugates
Sucrose-water mixtures of 40-75% w/w sucrose at variable temperature were used for producing a lifetime calibration set for sulfo-Cy3-HEWL as a function of viscosity; values of viscosity for different sucrose solutions were taken from literature. 14 In the case of BODIPY-HEWL conjugates, water-glycerol mixtures were used for calibration and viscosity values of these mixtures were calculated using the published empirical formula. 15 Time resolved fluorescence measurements was used to produce lifetime/viscosity calibrations of these mixtures as described below (in 'Time resolved measurements' section).

Blood clotting
For whole blood clot experiments, fresh venous blood was acquired by venepuncture from a consenting adult volunteer. 125 l of anticoagulant was added per millilitre of blood and the aliquots were stored at 4°C until required, within 48h.

SH-SY5Y cells obtained from ECACC (European Collection of Cell Cultures) were cultured in
Dulbecco's modified Eagle's medium (Sigma-Aldrich) which was supplemented with 10% Foetal Calf Serum. Cells were grown in T25 flasks for 4-5 days before being passaged after reaching

Transmission electron microscopy
The solution of 10 M Aβ(1-42) was prepared as described above and 4 μl aliquots were placed on Formvar/carbon coated 400 mesh copper grids (Agar Scientific) at the progressive time points to monitor aggregation. The Aβ(1-42) was allowed to adsorb onto the grid for two minutes before being blotted dry. The grid was then washed with 4 μl milliQ-filtered water and again blotted dry before being negatively stained with 4 μl of 2% w/v uranyl acetate for two minutes. The uranyl acetate was blotted and then left to air dry before being imaged using a JEM-1400Plus 120 kV transmission electron microscope (JEOL, USA). All images were acquired with a GATAN One View camera.

Absorption and fluorescence spectra
Absorption spectra were measured using an Agilent 8453 UV-Vis spectrophotometer.
Fluorescence spectra were measured using a FluoroMax4 spectrofluorimeter (Horiba Scientific) with a Xenon lamp as an excitation source. Spectra were corrected for wavelengthdependent efficiency of the light source and sensitivity of the detector. Quartz cuvettes with 1 cm path length were used in all measurements.
For the measurement of Thioflavin T fluorescence intensity, 10 µM Aβ(1-42) was allowed to assemble at 37°C in a cuvette inside the spectrometer (Horiba Scientific), fluorescence was excited at 430 nm and emission was collected at 480 nm, with both slits set to 5 nm.

Data analysis and representation
Multi-exponential fitting was done in SPCI software (Becker-Hickl) using the nonlinear least squares method and reconvolution algorithm for finding the best fit. Goodness of fit was judged by the χ 2 value and randomness of residuals. Decay models of fluorescence for all conjugated dyes were judged to be bi-exponential and followed the equation where I is fluorescence intensity, t is time, and αi are the amplitudes and τi the fluorescence lifetimes of the n exponentially decaying components. Data was further processed with OriginPro 8.6.
Mean fluorescence lifetime was calculated according to the equation Phasor analysis was chosen for data representation. This method does not require nonlinear multiexponential fitting and can be performed without prior knowledge of the number of components in the decay; it is highly suitable for large data sets and the visualisation of complex processes.
In this technique, fluorescence decays collected at a single frequency are Fourier transformed and their imaginary parts are plotted against the real parts in a so-called phasor plot. 12 The real (g) and imaginary (s) components are calculated as follows: where ω is the angular repetition frequency (2π x 80 MHz), t is time and I(t) is the measured fluorescence decay. The resulting vectors (g, s) are called phasors. Phasors of monoexponential decays lie on the "universal circle", a semi-circle centred at [0.5, 0], while phasors of multi-exponential decays lie on the inside of the semi-circle. Phasors corresponding to lifetime τ→0 would be found at [1,0]; with increasing lifetime the phasors move anticlockwise along the universal circle.
Phasor analysis was performed using an in-house MATLAB R2012a code. In our measurements, all phasors were corrected using an instrument response function.

General Materials and Methods
The manipulation of all air and/or water sensitive compounds was carried out using standard The synthesis of BODIPY-NHS started from the 4-(4-formylphenoxy)butanoic acid 1. 16 It is worth noting that the direct transformation of compound 1 into rotor BODIPY-NHS was hampered by the lack of solubility of derivative 1 in pyrrole. This problem was overcome by converting 1 into the corresponding silyl protected ester 2 in 60% yield. Therefore, BODIPY-COOH was obtained via condensation of compound 2 with an excess of pyrrole to give the corresponding dipyrromethane, which was further reacted with DDQ and followed by treatment with BF3·OEt2. 17 The silyl protecting group was removed during the work-up treatment. After purification by column chromatography, BODIPY-COOH was isolated in 22% yield from aldehyde 2. Finally, BODIPY-NHS was obtained in 55% yield by coupling of the carboxylic acid group of BODIPY-COOH with N-hydroxysuccinimide in the presence of DCC. 18 Compound 2. tert-Butyl(chloro)diphenylsilane (3.6 g, 12.9 mmol) was added to a mixture of 4-
The organic extracts were dried over anhydrous MgSO4, filtered and evaporated. The excess pyrrole was removed using high vacuum to give the dipyrromethane as a dark viscous oil. The crude dipyrromethane was purified by flash chromatography (2:1 CH2Cl2:petroleum ether) to give a green viscous oil. Yield: 2.5 g (90%). The corresponding dipyrromethane (1.2 g, 2.1 mmol) was dissolved in dry CH2Cl2 (100 mL) and DDQ (1.5 g, 6.6 mmol) was added under N2 atmosphere. The reaction mixture was stirred at room temperature shielded from light for 1 h.
Then, Et3N (3 mL, 21.3 mmol) was added, followed immediately by the addition of BF3·(OEt2)2 (3 mL, 24.4 mmol) and the reaction mixture was stirred at room temperature overnight. The