THREE-STEP RESONANCE ENERGY TRANSFER IN INSULIN AMYLOID FIBRILS

Uliana Tarabara1*, Mykhailo Shchuka1, Kateryna Vus1, Olga Zhytniakivska1, Valeriya Trusova1, Galyna Gorbenko1, Nikolai Gadjev2, Todor Deligeorgiev2 1Department of Medical Physics and Biomedical Nanotechnologies, V.N. Karazin Kharkiv National University 4 Svobody Sq., Kharkiv, 61022, Ukraine 2Faculty of Chemistry and Pharmacy, Sofia University, “St. Kliment Ohridski” 1, blv. J. Bourchier, Sofia, 1164, Bulgaria *E-mail: uliana.tarabara@gmail.com Received 15 October 2019, revised November 6, 2019; accepted November 11, 2019

During the past decade a phenomenon of the Fӧrster resonance energy transfer (FRET) has emerged as an extremely useful tool for retrieving information about proximity relationships and structural dynamics of biological macromolecules and their assemblies [1][2][3]. Due to a strong dependence of the energy transfer efficiency on the donor-acceptor distance, FRET is particularly useful while determining the intra-and intermolecular distances on a nanometer scale [4,5]. Although most FRET studies involve analysing a conventional one-step energy transfer, a multistep FRET (msFRET) has been attracting much attention in recent years [6][7][8][9][10], inspired by the natural photosynthetic systems containing several lightharvesting complexes that efficiently transfer the absorbed energy between a number of chromophores [11,12]. The energy transfer within multiple chromophore systems usually follows a cascade route, moving from an initial donor chromophore through the intermediate donors/acceptors onto a final acceptor chromophore [6][7][8][9][10]13]. The multistep FRET offers several advantages over the one-step FRET: i) a higher efficiency of long-range transfer [14]; ii) a larger Stokes shift [13,15]; iii) the possibility to monitor inter-and intramolecular interactions beyond the range 1-10 nm [16]; and iv) an extended excitation wavelength range for fluorescence lifetime measurements [14]. As a result, the cascade, or multistep FRET appeared to be especially useful in developing molecular photonic wires and light harvesting systems [6][7][8]17]. Most of such artificial systems are devised through synthesizing the arrays of covalently linked chromophores with a specific design to ensure large collection efficiencies, as well as fast and efficient energy migration. By analogy with the natural antenna, the majority of these systems are based on the porphyrin pigments [13,15,18,19]. Another application of msFRET involves the DNA photonic wires self-assembled around CdSe/ ZnS semiconductor quantum dots acting as a nanoscaffold and a FRET donor to a series of DNA-intercalating dyes [17]. More importantly, the msFRET systems are effectively used in biosensors for sensitive detection of multivalent complexation [20], protein labeling [15], genotyping of single nucleotide polymorphism [21], DNA sequencing [16], estimating the stoichiometry of protein complexes [22], determination of the tumor necrosis factor [23] and analysis of multiprotein interactions in living cells [24], to name only a few.
The main prerequisites for FRET include i) the overlap between emission spectrum of a donor and absorption spectrum of an acceptor and ii) the donor-acceptor separation falling in the range 0-10 nm. Each donor-acceptor pair can be characterized by the three parameters: t k , the energy transfer rate; r , the distance between chromophores; and 0 R , the Förster radius, related by the equation [27]: where D  is the donor fluorescence lifetime in the absence of acceptor. The energy transfer efficiency E , i.e. the probability that the excited donor will transfer energy to the ground-state acceptor, can be written as: The Förster radius, i.e. the donor-acceptor distance at which the energy transfer efficiency equals 50%, is given by: is the unit vector drawn from the donor to the acceptor [28]. The orientation factor may take a value from 0 to 4, but usually 0 R is calculated with  is the main limitation of the FRET technique resulting in the distance estimation error up to 35%) [28][29][30]. This problem can be partly circumvented through narrowing the 2  limits by measuring the anisotropy of donor and acceptor [28,29,31]. In the present study, the FRET system consisting of four chromophores is considered. There are many scenarios for this system and the two most important of them are illustrated in Fig.1. Let us designate the chromophores 1, 2, 3, and 4 as D1, D2, D3, and A3, respectively. In the case A (Fig. 1A), it is shown that the total three-step FRET efficiency (E) measured from the enhancement of the A3 emission is the product of efficiencies for each of FRET steps [33]: 12 23 34 where 12 E , 23 E and 34 E are energy transfer efficiencies from D1 to D2, from D2 to D3 and from D3 to A3, respectively. In the case B (Fig. 1B), the concentration of chromophores in the excited state can be described by the following differential equations: The energy transfer efficiency measured through monitoring the increase in acceptor fluorescence is given by [33]: (10) and (12) one obtains: Using the formula for the energy transfer efficiency within one donor-acceptor pair (Eq. 11), the E value in the threestep FRET can be written as follows: 12 23 34 13 34 12 24 14 where ij E is the energy transfer efficiency from the donor i to the acceptor j in the presence of the parallel energy transfer from the donor i to other acceptors.

EXPERIMENTAL SECTION
Materials Bovine insulin, dimethyl sulfoxide (DMSO), Tris, thioflavin T (ThT) and phosphotungstic acid hydrate for electron microscopy were purchased from Sigma. 4-dimethylaminochalcone (DMC) was from Signe (Latvia). The squaraine dyes SQ1 and SQ4 were synthesized in the University of Sofia, Bulgaria. All other reagents were used without further purification.

Preparation of working solutions
The insulin stock solution (10 mg/ml) was prepared in 10 mM glycine buffer (pH 2.0). The reaction of the protein fibrillization was conducted at 37 °C in the above buffer under constant agitation on the orbital shaker. The kinetics of amyloid formation was monitored using the Thioflavin T assay [34]. Hereafter, the fibrillar protein and its non-fibrillized counterpart (the insulin solution in glycine buffer that was not subjected to agitation) are denoted as InsF and InsN, respectively.
The dyes stock solutions were prepared in DMSO (SQ1 and SQ4) and ethanol (DMC), while ThT was dissolved in 10 mM Tris buffer (pH 7.4). The fluorimetric measurements were carried out in 10 mM Tris-HCl buffer (pH 7.4).
For the transmission electron microscopy assay, a 10 µl drop of the protein solution was applied to a carbon-coated grid and blotted after 1 min. A 10 μl drop of 1.5% (w/v) phosphotungstic acid solution was placed on the grid, blotted

Spectroscopic measurements
The absorption spectra of the examined dyes were recorded with the spectrophotometer Shimadzu UV-2600 (Japan) at 25 ºC. The dye concentrations were determined spectrophotometrically using the extinction coefficients  M -1 cm -1 for DMC, SQ1, SQ4 and ThT, respectively. Steady-state fluorescence spectra were recorded with RF6000 spectrofluorimeter (Shimadzu, Japan). Fluorescence measurements were performed at 25 °C using 10 mm pathlength quartz cuvettes. Fluorescence spectra were recorded within the range 460-820 nm with the excitation wavelength 440 nm. The excitation and emission slit widths were set at 10 nm.
The efficiency of energy transfer was determined from the quenching of the donor fluorescence in the presence of acceptor [29]: ( 1 5 ) where D I , DA I , are the donor fluorescence intensities in the absence and in presence of the acceptor, respectively. The donor fluorescence intensities measured in the presence of acceptor were corrected for inner filter effect using the following coefficients [29]: A are the acceptor optical densities at the donor excitation and emission wavelengths, respectively. The critical distance of energy transfer was calculated from eq. (4) using the Mathcad 15.0 software.

Molecular docking study
The molecular docking was carried out to ascertain the putative sites for the dye binding to insulin fibrils. The model of fibrillar insulin was taken from http://people.mbi.ucla.edu/sawaya/jmol/fibrilmodels/ [35]. The structures of the dyes were optimized using the semiempirical method PM6 (MOPAC2016 version18.012L) [36]. The top 10 conformations obtained with the PatchDock algorithm were then refined by the FireDock software [37]. The docked complexes were visualized by the Visual Molecular Dynamics (VMD) software.

Quantum-chemical calculations
Using the MOPAC2016 software, the geometry optimization of the dye conformations was performed, followed by the calculation of the quantum-chemical characteristics, such as: the solvent-accessible area ( CA ); molecular volume ( CV ); energy of the highest occupied ( HOMO E ) and lowest unoccupied ( LUMO E ) molecular orbitals; molecular length ( L ), height ( H ) and width (W ); ground state dipole moment ( g  ) and molecular weight ( . . M wt ). The LogP , a compound lipophilicity, was obtained using the ALOGPS 2.1 program (http://www.vcclab.org/lab/alogps/) [38]. All the calculated parameters are presented in Table 3.

RESULTS AND DISCUSSION
The ensemble of four dyes including the classical amyloid marker ThT (donor D1), chalcone dye DMC (acceptor A1 for ThT and donor D2), squaraine dyes SQ4 (acceptor A2 for DMC and donor D3) and SQ1 (acceptor A3) were selected to study the msFRET applicability to amyloid detection based on our previous results [25]. At the first step of the study we addressed the question of how the fluorescence of the terminal acceptor (SQ1) is sensitized by the other donors/acceptors of the energy transfer chain. To this end, the fluorescence spectra of the SQ1protein mixtures were measured with the excitation wavelength 440 nm (Fig. 4) that was chosen based on the spectral characteristics of the primary donor, ThT. The fluorescence intensity of SQ1 in the emission maximum (680 nm) did not exceed 1000 a.u. in the presence of fibrillar insulin (InsF). The addition of SQ4 (D3) into this system led to some decrease in SQ1 (A3) fluorescence, as well as to the appearance of the SQ4 fluorescence band. This implies that the expected enhancement of acceptor fluorescence does not occur at this stage, presumably because of relatively low fluorescence signal from SQ4 under the employed experimental conditions. At the same time, no SQ1 and SQ4 fluorescence was observed in the presence of the control protein (InsN). However, the subsequent addition of ThT resulted in the pronounced fluorescence increase in InsF, indicating that the energy is transferred from ThT to the squaraines. There are two possibilities by which such energy transfer may occur, the parallel pathway (ThT→SQ4 and ThT→SQ1) and the sequential one (ThT→SQ4→SQ1). Moreover, when the FRET chain was complemented by DMC serving as a bridge between ThT and SQ4, the squaraine fluorescence showed a further enhancement increasing with the elevation of DMC concentration (Fig. 4A,B). It is interesting to note that squaraine peaks emerging in the presence of DMC were observed also for InsN with SQ4 fluorescence signal being higher than that of SQ1, while in the absence of the mediator SQ1 and SQ4 fluorescence was negligibly small. The excitation spectra measured at the emission wavelength 720 nm at the highest DMC concentration also are indicative of the marked difference between the InsF and InsN (Fig. 4C). The squaraine peaks are clearly resolved in these spectra at 637 nm for SQ4, and 676 nm for SQ1, in the presence of the fibrillar insulin, while small hypsochromic shifts ca. 8 and 3 nm for SQ4 and SQ1, respectively, along with significantly lower fluorescence intensities were observed for the control non-fibrillized protein.  The ratio of the signal amplification in the presence of the fibrillar protein to that in the control showed somewhat decrease with increasing the DMC concentration (Fig. 5C). However, at all mediator concentrations, the dye fluorescence enhancement was more than 3 times stronger in the presence of fibrillar insulin with respect to the control protein for both SQ1 and SQ4.  At the next step of the study, the ThT-protein mixtures were consecutively titrated with DMC, SQ4 and SQ1.The donor fluorescence gradually decreased with increasing acceptor concentration for each FRET step in the presence of both fibrillar and control protein, except for the ThT-DMC pair in InsN, where the donor and acceptor fluorescence increased simultaneously due to superposition of the DMC and ThT emission spectra (data not shown). The efficiencies of energy transfer for the three donor-acceptor pairs were calculated from the quenching of the donor fluorescence. As illustrated in Fig. 6, FRET efficiencies in the presence of fibrillar insulin increased with increasing acceptor concentrations by the factors of 3, 2.4 and 7.5 for the pairs ThT-DMC, DMC-SQ4 and SQ4-SQ1, respectively. The largest values of FRET efficiency were observed for the pair DMC-SQ4 (E> 60%). Remarkably, the energy transfer efficiencies in the presence of fibrillar insulin were ~ 12 times greater and ~1.5 times lower than those for the control protein for the pairs D2-A2 and D3-A3, respectively. A more pronounced FRET observed in the control protein for the pair SQ4-SQ1 as compared to InsF did not have a critical influence on the resulting transfer efficiency, since at the first step of cascade FRET the energy was not transferred in InsN and at the second step the efficiency was lower than that in InsF.
As seen from Eqs. (3) and (4), the efficiency of energy transfer depends on the distance between donor and acceptor and the Förster radius which, in turn, is determined by the overlap between the donor emission and acceptor absorption spectra, the donor quantum yield and the acceptor extinction coefficient, the refractive index of the medium and the orientation factor. The overlap integrals, the Förster radii and the donor-acceptor separations were calculated for the pairs ThT-DMC, DMC-SQ4 and SQ4-SQ1 taking the isotopic value of the orientation factor ( 2 2 / 3    1 7 ) where s Q is the quantum yield of the standard, d A and s A are the optical densities at the donor excitation wavelength, d S and s S are the areas under the fluorescence bands, d n and s n are the refractive indexes of the medium for the donor and standard, respectively. Notably, the binding of ThT to fibrillar insulin resulted in the increase of the dye quantum yield by more than two orders of magnitude (Table 1). The normalized emission and absorption spectra of the three donor-acceptor pairs are depicted in Fig. 7, while the overlap integral value (J) evaluated by the numerical integration are presented in Table 2. The pair SQ1-SQ4 demonstrated the greatest J value due to a small shift (~32 nm) between SQ4 fluorescence and SQ1 absorption maxima, and a high extinction coefficient of SQ1.

EEJP. 4 (2019)
The Förster radius, a characteristic of each donor-acceptor pair, was calculated from equation (4). The largest value of the Förster radius ( The distinctions between the fibrillar and control proteins appeared to be the most pronounced in the 3D fluorescence spectra. As illustrated in Fig. 8A, in the system consisting of ThT, DMC and SQ4 bound to InsF, the strongest fluorescence centered around ~ 650 nm corresponds to SQ4; the spot located above the excitation wavelength 550 nm is related to the direct excitation of SQ4, while the bottom spot centered at λ Ex ~ 440 nm originates from the energy transfer ThT→DMC→SQ4. A residual ThT fluorescence is observed at λ Ex ~ 440 nm and λ Em ~ 580 nm. The addition of SQ1 into the above ternary system resulted in the appearance of a new spot centered at λ Em ~ 685 nm and λ Ex ~ 440 nm, and in the decrease of SQ4 signal (Fig. 8C), suggesting that the energy is transferred from ThT to SQ1 via SQ4. As seen in Figs. 8B and 8D, the intensity of the analogous patterns is substantially lower in InsN compared to InsF. The barely noticeable spots centered at the SQ4 (Fig. 8B) and SQ1 (Fig. 8D) emission maxima at λ Ex ~ 440 nm were observed, while the higher emission intensities observed at λ Ex > 550 nm are related to the direct excitation of the fluorophores. In the last step of our investigation, the simple docking studies were performed to provide additional structural characterization of the dye-fibril complexes. As seen in Fig. 9A, ThT, DMC and SQ4 tend to associate with the 4-5 residues of the L17 ladder of the insulin fibril protofilament located on the dry steric zipper interface, suggesting the predominant role of hydrophobic and van der Waals dye-protein interactions. A similar binding motif was previously proposed for ThT [42]. In turn, the bulky moieties of zwitterionic SQ1 seem to prevent its binding to the dry amyloid
surface, and thus, SQ1, being the most hydrophobic among the examined dyes (possessing the highest LogP value, Table 3) binds to the wet surface groove formed by the residues GLN15 and GLU17 of the insulin A-chain. In this case, the hydrophobic, aromatic and electrostatic forces may stabilize the dye-protein complex. Interestingly, the distance between fibril-bound SQ4 and SQ1 was about 3 nm, being close to the value ca. 3.9 nm obtained from the FRET efficiency (Table 2). However, precise identification of the fibril binding sites for the dyes seem to be complicated because the rigid docking algorithms used here are based on the shape complementarity principles (PatchDock) and perform only the protein side-chain optimization (FireDock). In turn, the ligand structure remains fixed during the docking process that may lead to the appearance of the false protein binding sites if the optimized dye geometry used for calculations differs from the real geometry. Thus, the discrepancies between ThT-DMC, and DMC-SQ4 distances obtained experimentally ( Table 2) and by the molecular docking (Fig. 9A) can be explained by the drawbacks of the docking technique. To  (Table 3). If SQ4 and SQ1 are associated with the two opposite sides of the fibril surface, the intermolecular distance will be about 3 nm. In turn, if ThT and DMC are bound to the L17 ladders available on the first and second β-sheet of the insulin protofilament, respectively, the distance between the two fluorophores will be about 2.5 nm (Table 2).
A B Fig. 9. Schematic representation of the energetically most favorable dye complexes with fibrillar insulin, obtained using PatchDock/ FireDock servers and visualized by VMD software.
The drawing method was set as Bonds and NewCartoon for the dyes and the protein, respectively. ThT, DMC, SQ4 are bound to the L17 ladder of the B chain, located at the dry steric zipper of the insulin fibril protofilament, while SQ1 is attached to the surface groove formed by the residues GLN15 and GLU17 of the A chain (A). The distance between SQ1 and SQ4 is about 3 nm (B).

67
Three-Step Resonance Energy Transfer in Insulin Amyloid Fibrils EEJP. 4 (2019) CONCLUSIONS To summarize, we have successfully designed the dye ensemble suitable for the three-step sequential FRET in the insulin amyloid fibrils. The following chromophores were recruited as the components of the fibril-scaffolded FRET chain: i) a primary donor, Thioflavin T, transferring its energy to the mediator dye, DMC, upon excitation at 440 nm; ii) a novel squaraine dye SQ4 accepting the energy from DMC; and iii) squaraine fluorophore SQ1 accepting the energy from SQ4, followed by the emission enhancement at 680 nm. The above cascade showed notable increase in the SQ1 fluorescence compared to that devoid of DMC. Despite the parallel FRET between the dyes cannot be excluded, the donor-acceptor distances estimated for each pair assuming the isotopic chromophore rotation appeared to be consistent with the intergroove separations in the core of insulin fibrils. The recovered distances fall in the range 2.4 -4.5 nm, suggesting different fibril binding sites for the dyes. The fact that no cascade FRET was observed in the non-fibrillized protein, highlights the importance of the dye association with the highly ordered amyloid structure for the energy transfer to occur in the chain ThT→DMC→SQ4→SQ1. These results may prove of importance in the development of the novel sensitive fluorescence approaches to amyloid detection and characterization, particularly, in vivo, by introducing the nearinfrared fluorophores to the cascade.