Computational Investigation of the Binding Dynamics of Oligo p-Phenylene Ethynylene Fluorescence Sensors and Aβ Oligomers

Abstract

Amyloid protein aggregates are pathological hallmarks of neurodegenerative disorders such as Alzheimer’s (AD) and Parkinson’s (PD) diseases and are believed to be formed well before the onset of neurodegeneration and cognitive impairment. Monitoring the course of protein aggregation is thus vital to understanding and combating these diseases. We have recently demonstrated that a novel class of fluorescence sensors, oligomeric p-phenylene ethynylene (PE)-based electrolytes (OPEs) selectively bind to and detect pre-fibrillar and fibrillar aggregates of AD-related amyloid-β (Aβ) peptides over monomeric Aβ. In this study, we investigated the binding between two OPEs, anionic OPE₁²⁻ and cationic OPE₂⁴⁺, and to two different β-sheet rich Aβ oligomers using classical all-atom molecular dynamics simulations. Our simulations have revealed a number of OPE binding sites on Aβ oligomer surface and these sites feature hydrophobic amino acids as well as oppositely charged amino acids. Binding energy calculations show energetically favorable interactions between both anionic and cationic OPEs with Aβ oligomers. Moreover, OPEs bind as complexes as well as single molecules. Compared to free OPEs, Aβ protofibril bound OPEs show backbone planarization with restricted rotations and reduced hydration of the ethyl ester end groups. These characteristics, along with OPE complexation, align with known mechanisms of binding induced OPE fluorescence turn-on and spectral shifts from a quenched, unbound state in aqueous solutions. This study thus sheds light on the molecular-level details of OPE-Aβ protofibril interactions and provides a structural basis for fluorescence turn-on sensing modes of OPEs.

INTRODUCTION

Alzheimer’s disease (AD) is the most prevalent neurodegenerative disease, with 50 million affected globally in 2017.¹ A projected 131.5 million worldwide will likely suffer from AD by 2050.¹ Diagnosis of AD occurs well after the onset of clinical symptoms, often postmortem. Intraneuronal neurofibrillary tangles and extracellular amyloid plaques are two hallmark lesions of AD. As the tangles and plaques are believed to be formed well before, perhaps by decades, the onset of neurodegeneration and cognitive impairment,^2,3 monitoring their formation is thus vital to understanding, detecting, and combating these diseases.

Amyloid plaques are large assemblies composed of fibrillar aggregates of amyloid beta (Aβ) peptides. Aβ is formed from the transmembrane amyloid precursor protein (APP) following cleavage by multiple secretases.⁴ While the precise function of Aβ is unclear, monomeric Aβ appears to be involved with neural memory and learning processes.^5,6 The aggregate-prone form of the protein results in Aβ fibrils containing an organized cross-β scaffold with extensive regions of non-polar residues.^7,8 Polar and charged amino acids in these aggregates are aligned on the periphery around a hydrophobic core. Recent evidence has shown that smaller, more soluble Aβ oligomers and protofibrils are the most neurotoxic aggregates and the primary cause of cognitive impairment.^4,9 In fact, it has been posited that an inverse correlation exists between oligomer size and toxicity.⁴ Selective detection of these smaller Aβ aggregates presents a significant challenge due to their heterogeneous and dynamic nature.^10–16 Although Aβ oligomers have the same residues, their conformations are more unstructured than fibrils,¹⁷ which could render a number of commonly used fibril binding dyes ineffective at detecting oligomers.

Fluorescent dyes with conjugated rod-like motifs such as thioflavin T (ThT)^12,18 and Congo Red (CR)^19,20 have been successfully used to detect Aβ fibrils but are limited by a single sensing mode and non-specific binding even to native proteins.^21,22 Oligomeric p-phenylene ethynylenes (OPEs), a class of versatile, multifunctioning compounds that have been found useful in antimicrobial^23–25 and sensing^26–29 applications, have shown promise at detecting fibrillar aggregates of model proteins and more recently, aggregates of disease-relevant Aβ, α-synuclein, and tau proteins.^30–32 These studies have identified two promising OPE compounds, one to two repeat units in size, with either cationic or anionic charged groups along the backbone and ethyl ester termini (Fig. 1). We recently showed that these OPEs can selectively and sensitively detect Aβ fibrils and oligomers.³³ Several factors have been hypothesized to contribute to OPE fluorescent sensing, including hydrophobic interactions, J-dimer formation, backbone planarization, and unquenching of the ethyl ester end groups upon binding.^30–35 However, these effects have not been studied and demonstrated at the molecular level.

Figure 1.

Structures of OPEs investigated in this study.

A number of computational approaches including molecular dynamics (MD) simulations and molecular docking have been used to investigate properties of a plethora of therapeutic and diagnostic molecules relevant to AD.^36–46 Both ThT and Congo Red have been studied and reported to favorably bind to Aβ protofibrils that result in the planarization of backbone ring groups.^{36–38,42,43,46} Combined MD and quantum mechanics calculations have also provided molecular scale details of binding and sensing mechanisms of other amyloid dyes such as oligo-thiophenes.^45,47

In this study, we explore the interactions of two oppositely charged OPE oligomers, OPE₁²⁻ and OPE₂⁴⁺ (Fig. 1), with two Aβ40 oligomers consisting of either 5 peptides (5-mer) or 24 peptides (24-mer) using atomistic, explicit solvent MD simulations. As this study is the first with OPEs and Aβ proteins using molecular dynamics, we selected an aggregate featuring the more common Aβ40 sequence⁴⁸ with the long-term goal of screening a library of amyloids including Aβ42 in future simulations. We also elucidate molecular-level details that are likely responsible for the enhanced fluorescence turn-on observed upon binding of OPEs to Aβ aggregates. Specifically, OPE binding sites were analyzed to determine the interactions that promote OPE association in addition to determining OPE self-assembly or complexation on the Aβ surfaces. Hypothesized modes of OPE-binding induced fluorescence turn-on were each examined to gain a detailed understanding of OPE binding as well as fluorescence turn-on mechanisms.

RESULTS AND DISCUSSION

OPEs readily bind to Aβ oligomers

Both cationic and anionic OPEs were observed to bind to the surfaces of the 5-mer (Systems I-A and I-B) and the 24-mer (Systems II-A and II-B) in all simulations in less than 100 ns. In general, binding occurred with complexation of OPEs on the surface of the protofibrils. We determined the amount of bound OPEs at both the ends and periphery of the 24-mer and found that the majority of the compounds found binding sites on the outer surface compared to the ends (Table S1). In addition, the OPEs formed complexes on the 24-mer readily. Representative snapshots of simulation trajectories for all systems are shown in Fig. 2 with charged OPE side pendant groups highlighted by yellow or blue spheres representing negative and positive charges, respectively. Aβ peptides are shown in ribbon representation with purple, orange, and gray denoting β-sheets, helices, and random coils, respectively. Binding was analyzed by contact point analysis between OPE and Aβ, wherein binding is defined when the distance between any atom of the OPEs and Aβ oligomer is within 4 Å. As shown in Fig. 3, binding occurred during the first 20 ns and the majority of OPEs stayed bound throughout the remainder of the trajectories. An OPE concentration of 10.2 mM (2 OPEs per Aβ peptide for a total of 10 OPEs) was used in Systems I-A and I-B, while a concentration of 8.74 mM was used in Systems II-A and II-B (2 Aβ peptides per OPE for a total of 12 OPEs). Note that these concentrations, roughly three orders of magnitude larger than the μM concentrations used in experiments, were chosen in order to increase sampling of binding sites in addition to allowing observation of any OPE self-assembly or complexation. This approach has been successfully implemented in prior studies by our group investigating the interaction between biflavonoid curcumin and Aβ⁴⁹ and in other similar computational studies investigating ligand binding to Aβ such as Congo Red.⁵⁰

Figure 2.

Snapshots of bound and non-bound OPEs at 100 ns for oligomers (Systems I-A and I-B) and 24-mers (Systems II-A and II-B). OPEs are shown in stick representation with yellow and blue spheres indicating anionic and cationic charge groups, respectively.

Figure 3.

Representative contact point profiles, defined as less than 4 Å between any atom of OPE and Aβ, for singly bound OPE₁²⁻ (A) and OPE₁²⁻ bound as complexes (B) in System II-A. Snapshots from MD trajectories at 0 and 100 ns are shown with the Aβ 24-mer in ribbon representation.

Simulations showed that both cationic and anionic OPEs bound to the anionic Aβ 5-mer. Out of 10 OPEs positioned around the 5-mer, 10, 8, and 4 of OPE₁²⁻ (System I-A) and 5, 6, and 7 of OPE₂⁴⁺ (System I-B) became bound in each of the three trajectories. Of the 40 bound OPEs, 7 were single OPEs, and 33 were complexes consisting of up to 5 OPEs. OPE₂⁴⁺ predominantly formed dimers while OPE₁²⁻ assembled into larger complexes including tetramers and pentamers. Smaller OPE₂⁴⁺ complexes could be the result of high electrostatic repulsion since each molecule has a net charge of +4. The small size of the 5-mer could have also limited the amount of OPE binding, especially for the larger OPE₂⁴⁺, which is 33 Å in length, whereas the dimensions of the 5-mer were 19 × 38 × 49 Å.

In both Systems II-A and II-B, more OPEs (11 out of 12) became bound to the larger Aβ 24-mer either as single molecules or as complexes. The increase in total number of bound OPEs to the 24-mer compared to the 5-mer was likely a result of the larger surface area (33.1 nm² for the 24-mer versus 10.2 nm² for the 5-mer) accessible to binding. Binding also occurred frequently on the edges and ends of the 24-mer with few OPEs also found in the β-sheet regions. The OPE complexes were predominantly dimers, and these were formed either on the protofibril surface after one OPE became bound or after the OPEs self-assemble first before binding. Similar to binding to the 5-mer, OPEs bound rapidly, in less than 20 ns, on the 24-mer, and in most cases were bound for the entire 100 ns. Contact analysis (Fig. S2) of System II-A showed some instances in which binding was delayed until after the 50 ns due to OPE₁²⁻ self-assembly occurring in solution first. Once bound, the majority of OPEs remained in close contact with the 24-mer. However, a dimer in System II-A drifted away from the surface before binding again at 90 ns. We extended the simulation time of this particular system to 200 ns and found that this OPE dimer remained bound to the 24-mer thereafter.

Charged residues influence OPE-Aβ interactions, but non-polar residues primarily mediate the binding process

To characterize the OPE binding sites, we explored the residue distribution surrounding the bound OPEs by determining the number of amino acids within 4 Å from OPE at the 100 ns time point. Fig. 4 shows the resulting residue counts and percentages of amino acid categories (polar uncharged, cationic, anionic, and non-polar) for singly bound OPEs and the innermost OPE of complexes bound to the Aβ 5-mer. Several trends in the binding site residue compositions were observed. OPE₁²⁻ and OPE₂⁴⁺ binding sites have significant percentages of non-polar residues on the 5-mer, 68% and 77%, respectively. Residue counts and percentages for the 24-mer (Fig. S3) show a similar trend with non-polar amino acids being most prevalent at binding sites. This is not surprising as Aβ oligomer contains 70% non-polar residues on its surface and the PPE backbone of the OPEs is hydrophobic. The most significant difference between the two OPEs is in the charged residues at their binding sites. In the case of OPE₁²⁻, the ratio of anionic to cationic residues within 4 Å is approximately 0.25 while the same ratio for the cationic OPE₂⁴⁺ is 3. This result clearly shows that favorable electrostatic interactions play a role in OPE binding, and although each Aβ peptide has a net −5 charge, the OPEs preferentially bind in regions of opposite charge on the protein surface.

Figure 4.

Number, percentages and types of amino acids (non-polar, polar uncharged, anionic, and cationic) on the 5-mer within 4 Å of each particular bound OPE₁²⁻ (A) or OPE₂⁴⁺ (B) along the y-axis at 100 ns.

To compare the binding site residues of singly bound and complex bound OPE₁²⁻ over time, additional residue counts were carried out in 10 ns intervals after the mid-point of the trajectories (50 ns) of System II-A. As anticipated from visualizing OPE binding along the trajectories, these results (Fig. 5) revealed varying levels of change in the residue distribution. The majority of compounds bound tightly to the 24-mer, especially when complexed with other OPEs (Fig. 5B). On average, the number of cationic and non-polar residues within 4 Å of the OPE complex binding site changed by 1 to 2 amino acids between time points (Fig. 5B), compared to a single bound OPE on the 24-mer edge which increased by up to 6 amino acids from 70 to 80 ns (Fig. 5A). This behavior also indicates the difference in the binding sites as OPE complexation buries some OPEs underneath others, restricting their movements (Fig. 5B). In contrast, the single OPE bound to the 24-mer end (Fig. 5A) moves more freely since there is no other OPE present at this site. This behavior can also be seen in Fig. 3 where the contacts between an OPE₁²⁻ bound at the end of the 24-mer fluctuated whereas contacts between a complex bound OPE₁²⁻ were less dynamic. We also tracked specific amino acid residues to further analyze changes in binding sites over time. In the case of the complex bound OPE, roughly 93% of the residues were conserved throughout the final 50 ns while only ~68% were conserved during the same interval for the singly bound OPE investigated in Fig. 5.

Figure 5.

Number of residues over time for a singly bound OPE₁²⁻ (A) and a complex-bound OPE₁²⁻ (B) over the final 50 ns of system II-A.

In order to elucidate the energetics of OPE binding, we analyzed the final 15 ns of each trajectory using MM-GBSA calculations. Binding energies, ΔG_Bind, for complexed OPEs were obtained by treating the entire complex as a “ligand” that binds to the protein receptor. ΔG_Bind values reported in Table 1 were calculated for single and complexed OPEs bound to Aβ 5-mers (averages of all bound OPEs from three trajectories) and Aβ 24-mers (averages of all bound OPEs in a single trajectory). Errors reported are standard deviations of ΔG_Bind values averaged over the final 10 ns of each trajectory. Statistical analysis of average ΔG_Bind values of single and complex bound OPE revealed no significant difference between the two populations.

Table 1.

Average binding energies for singly bound or complexed OPEs in kcal/mol.

	OPE12−		OPE24+
	single	complex	single	complex
small protofibril	−25.3 +/− 7.2	−36.0 +/− 13.6	−36.6 +/− 7.7	−34.2 +/− 10.1
large protofibril	−31.7 +/− 6.6	−39.3 +/− 9.5	−33.1 +/− 6.2	−32.9 +/− 7.7

Energetics of OPE binding were further broken down to identify different binding contributions (Table S2). Interestingly, the electrostatic components (ΔG_EL) of the binding energies are always negative (favorable) in the case of OPE₂⁴⁺ and mostly positive for OPE₁²⁻. Since the charge of the Aβ peptide is negative, it is not surprising that there is an unfavorable energetic effect associated with binding of the anionic OPE₁²⁻. The generalized Born solvation terms, ΔG_GB, show opposite signs to ΔG_EL. This means that for OPE₁²⁻, introduction of the dielectric environment screens the electrostatic repulsion. However, the favorable electrostatic interactions between OPE₂⁴⁺ and Aβ is also reduced by the dielectric environment. Nevertheless, the competition between electrostatic and solvation components are always overcome by van der Waals (VDW) effects and favorable reductions in solvent accessible surface area from OPE binding. The net result is favorable ΔG_Bind for all bound OPE₂⁴⁺ and OPE₁². All of the binding energy calculations show a multitude of favorable sites on the Aβ 5-mers and 24-mers that promote OPE association and complexation.

OPE binding causes backbone restriction and reduction in OPE end group hydration

Backbone planarization due to rotational restriction is the proposed mechanism for fluorescence enhancement of well known “rod-like” sensors including ThT and Congo Red.^51,52 OPE sensing of amyloids has been attributed in-part to backbone planarization, although other factors also contribute to binding-induced fluorescence enhancement and spectral shifts.^30,32,33 In this study, backbone planarization was assessed by analyzing the pseudo-dihedral (p-dihedral) angles between the phenyl ring groups of free OPE₁²⁻ in solution, singly bound, or complex-bound OPE.⁵³ Fig. 6 shows the distribution of p-dihedral angles from −180 to 180° over the last 50 ns of simulation for System II-A. The two p-dihedrals of OPE₁²⁻ are shown by red and black dots and there are clear differences in their distributions in the three OPE states. For free OPE in solution (top of Fig. 6), there is no preferred backbone rotational conformation where both dihedrals sampled the full range of values. Upon binding as single molecules, the OPE backbone became somewhat restrained as shown by the middle panel of Fig. 6 where p-dihedral values were somewhat clustered, but still sampled the full range. Analysis of additional singly bound OPEs show different extents of dihedral angle clustering; OPE bound in close proximity to the loop-regions on the 24-mer surface exhibited a more planar conformation (Single 3 in Fig. S4), while OPEs bound to other binding sites where the OPEs extend away from the protofibril exhibited more unrestricted rotation (Single 1 in Fig. S4). In contrast, when the OPEs were bound in a complex, especially those in trimers or larger complexes, there was significant restraint of backbone rotation resulting in planarization. This pattern was observed in the bottom panel of Fig. 6 in which the OPE bound in a tetramer exhibited significant clustering of p-dihedrals at 0 degree, which corresponds to a planarized backbone conformation. We performed similar analysis of a free and complex-bound OPE₂⁴⁺ (data not shown) on the 24-mer and found that restriction to backbone restriction occurred as well, albeit to a lesser extent than OPE₁²⁻. As OPE₂⁴⁺ only formed dimers in our simulations, the restriction in p-dihedrals is less pronounced compared to OPE₁²⁻. This is consistent from our previous quantum mechanics study³⁴ which found that the delocalization length for OPEs is approximately 3 phenyl rings long. The rotational barriers about the p-dihedrals for OPE₂⁴⁺ are therefore lower than that for OPE₁²⁻. This means full planarization of the five-ring backbone of OPE₂⁴⁺ is unlikely.

Figure 6.

Pseudo-dihedrals between each pair of phenyl rings (shown in red and black dots) for last 50 ns of free (top), singly bound (middle), and complexed OPE₁²⁻ (bottom) bound to large 24-mer. Schematic showing atoms chosen for dihedral analysis (blue circles) and corresponding angles sampled are given on the right.

In addition to the p-dihedral analysis, we also determined the overall planarity of OPE₁²⁻. The total planarity, P, was calculated by summing all dihedrals for each time point after binding had occurred⁵³ (Equation 1):

$$P=\sum _i\frac{\Vert {\theta }_i\left|-90\right|}{90}$$

A P value of 2 indicates a fully planar backbone and a value of 0 indicates a backbone completely out of plane. In order to compare the distributions of P values for the free and bound OPE₁²⁻, box and whisker plots were generated (Fig. 7). The complex-bound OPE was in a more planar conformation with most of the P values falling between 1.5 and 2 compared to a single bound molecule with P values predominantly between 1 and 1.5. By comparison, the free OPE in solution was the least planar with the majority of P values lying between 0.5 and 1.5. Fig. S5 shows additional plots for other Aβ bound OPE₁²⁻ compared to a free OPE₁²⁻. Similar trends were observed in these cases but the very obvious differences in planarity between single and complex bound OPEs is less pronounced in these trajectories. This is largely due to the variability in the binding sites of the OPEs, each resulting in distinct and unique patterns of backbone planarization. Certain OPE binding and complexation sites resulted in the OPEs binding more tightly to the surface of the protein, and in conformations that restrict rotations of the backbone.

Figure 7.

Box and whisker profiles showing distribution of planarity values for free (blue), singly bound (orange), and complexed (gray) OPE₁²⁻ on 24-mer with a value of 2 indicating a fully planar backbone.

Binding induced fluorescence turn-on has additionally been attributed to unquenching of ethyl ester end groups on the OPEs.^30,32,54 Earlier experimental and theoretical work has been carried out to understand why the ethyl ester moieties may cause this unquenching behavior. Interfacial water contributions were evident after fluorescence measurements were obtained for ethyl ester terminated OPEs in both water and deuterium (absent of any protein).⁵⁵ MD simulations from this study also showed increased hydrogen bonding between water molecules upon addition of the end groups, which likely are involved in red-shifted fluorescence signatures observed when OPEs bind to protein aggregates.⁵⁵ It is also worth noting that no other OPE end groups have caused quenching of the excited singlet state of OPEs. To explore evidence of this effect during binding of OPEs to the 24-mer in this study, we determined the total number of water molecules within 3.4 Å of the ethyl ester terminated ends of the OPEs over time for free, singly bound, and complex bound OPE₁²⁻ in System II-A (Fig. 8). Representative snapshots (green boxes in Fig. 8) are shown for the 0 and 100 ns time points. For free OPE, the number of waters surrounding the end groups remained stable at approximately 15. Upon binding as a single molecule on the end of a 24-mer (Fig. 8B), one OPE₁²⁻ showed a slight decrease in the number of water molecules hydrating the end groups. This particular binding site still allowed the ethyl ester termini to be mostly exposed to the water environment. OPE binding as complexes significantly reduced the number of waters around the end groups during the first 10 ns of simulation (Fig. 8C). Our simulation results thus confirm that OPE binding to Aβ oligomer reduces the hydration of the ethyl ester end groups, and support binding-induced unquenching as a mode of fluorescence enhancement for OPE sensing.

Figure 8.

Number of water molecules over time for free (A), single bound (B), and complex-bound (C) OPE₁²⁻ in system II-A. Green insets show trajectory snapshots at the beginning (0 ns) and end (100 ns) of the trajectories with water molecules within 3.4 Å of ethyl ester termini in red and white spheres.

MM-GBSA calculations were performed to gain insights into binding energetics between OPEs and Aβ including detailed breakdown of electrostatic and van der Waals components (Fig. 9). Binding energies of a few other compounds are also plotted for comparison. As shown, the van der Waals (VDW) component of the OPE-oligomer binding free energy (ΔG_VDW) as well as overall binding energy (ΔG_bind) values are all negative, indicating favorable binding. A number of simulation studies have found that favorable binding of ThT and Congo red (CR) to Aβ and related amyloids is largely a result of the strong VDW interactions between the conjugated backbone and the protein.^56–60 Average ΔG_VDW values calculated in this study for the binding of OPE dimers, trimers, and tetramer to Aβ were in the range of −22 to −70 kcal/mol, and is similar to ΔG_VDW values of curcumin binding to an Aβ 24-mer that ranged from −30 to −40 kcal/mol.⁴⁹ ΔG_VDW the OPEs are more negative (favorable) likely because of increased attractive forces of induced dipoles along the longer backbone of the OPEs compared to curcumin.

Figure 9.

ΔG_VDW components for dimers (black), trimers (red), and tetramers (green) of OPEs and curcumin (A).⁴⁹ Comparison of total ΔG_Bind of OPEs, curcumin, CR, and ThT based compounds (B).^36,37

ΔG_Bind values of OPE₁²⁻ and OPE₂⁴⁺ were −28.5 +/− 6.8 kcal/mol and −34.9 +/− 6.9 kcal/mol, respectively (Fig. 9), and they are more negative, indicating more favorable binding, than ΔG_Bind previously calculated for curcumin complex binding to an Aβ 24-mer⁴⁹ and those previously reported for ThT,³⁶ BTA-1 (a neutral ThT analog),³⁶ and CR.³⁷ While it is important to note that the fibril structures in the studies of ThT, BTA-1, and CR are not identical to the proteins in this study, all of them have a similar β-sheet enriched core and, in the case of ThT and BTA-1, an abbreviated version of Aβ40. Indeed, we have found experimentally that both OPEs bind with other Aβ40 fibril morphologies with higher affinities, or lower dissociation constants, compared to ThT.³³ Our simulation results showed that the charge moieties on OPEs and their locations influence binding behavior. The OPEs have two charge groups per subunit attached to alkyl side chains positioned away from the backbone. This flexible side chain structure allows for stronger OPE binding to sites with oppositely charged amino acids (Fig. 10). In contrast, charged groups on CR and ThT are directly attached to the backbone aromatic ring groups, which reduces the ability of the compounds to optimize charge interactions at binding sites.

Figure 10.

OPE₁²⁻ (A) binding to a cationic, lysine rich site (blue) and OPE₂⁴⁺ (B) binding to sites enriched with anionic glutamic and aspartic acid residues (red). OPE charge groups are shown as spheres with anionic and cationic moieties in red and blue, respectively.

It is evident from our results that there are two major contributions to OPE binding: (i) favorable electrostatic interactions between OPE and oppositely charged residues at protein binding sites, and (ii) VDW and hydrophobic interactions of the OPE backbone with the protein surface. These interactions resulted in fast and persistent binding of the OPEs to Aβ oligomers.

Our results provide insight into proposed OPE binding-induced fluorescence turn-on mechanisms, namely complexation, backbone planarization, and unquenching of ethyl ester end groups at the molecular-level. Distinct restrictions to backbone rotation were observed for both OPE₁²⁻ and OPE₂⁴⁺, especially in cases where complexation caused “sandwiching” of one compound between the Aβ peptides and other bound OPEs. This resulted in flattening of the aromatic rings of the OPE which extended the π-conjugation length of the segment chromophore.³⁴ Lindgren and coworkers have similarly used MD simulations in conjunction with experimental work to explore spectral properties including backbone planarity of a series of oligothiophenes and the implications in binding to insulin fibrils.⁶¹

Earlier work has shown unquenching effects when OPE compounds are in more non-polar environments such as methanol⁵⁴ and when complexed with detergents such as sodium dodecyl sulfate (SDS).³⁵ These observations led us to propose that OPE fluorescence enhancement from fibril binding could be due to unquenching of OPE ethyl ester end groups.³² This phenomenon was clearly observed in our MD simulations which showed reduced number of water molecules within the first solvation shell (3.4 Å) of the bound OPE₁²⁻ end groups compared to free OPEs, especially in the case of bound OPE complexes where the average number of waters dropped as much as 10 molecules per OPE after the first 10 ns of simulation.

The results from this work also enhanced our understanding of the different energetic components involved with OPE-Aβ interactions. Experimentally determined ΔG_Bind values from our previous work are approximately −10 kcal/mol for OPE₁²⁻ and OPE₂⁴⁺ compounds,³³ which is smaller in magnitude compared to the ΔG_Bind values of approximately −30 kcal/mol calculated in this study. Differences between experimental ΔG_Bind and MD values obtained through MM-GBSA calculations have been well-documented.⁶² For instance, differences of up to 30 kcal/mol between experimental and simulated ΔG_Bind values of various ligand binding with α-thrombin have been reported.⁶² MM-GBSA calculations are known to underestimate the stability of free ligand and receptor which in turn leads to higher energetic contributions and larger magnitude ΔG_Bind compared to experimental values. The discrepancies in OPE-fibril binding energetics could also be due to the differences in the solvent environment. Additionally, conformational entropy contributions are often absent in MM-GBSA calculations and can contribute to differences in ΔG_Bind values. However, a recent study on ThT and a potential PET probe, AZD2184,⁴² included this effect and found that ΔG_Bind of ThT to Aβ 5-mers became less negative (more unfavorable) by as much as 24 kcal/mol leading to a ΔG_Bind of −4.7 kcal/mol for one particular binding site.

CONCLUSIONS

Our simulations have provided molecular level insights into the binding of OPEs to Aβ aggregates. Both anionic and cationic OPEs bind to a number of sites on the two Aβ oligomers. OPEs bind as single molecules but also readily form complexes in a staggered, brick-type stacking configuration that is characteristic of J-dimerization⁶³ wherein the charge groups are positioned away from one another. While the binding sites of the OPEs are composed of predominately non-polar residues, oppositely charged amino acids are also a clear feature of the binding sites. Thus, OPE-Aβ oligomer binding is mediated by both hydrophobic and electrostatic interactions. Calculated ΔG_Bind values for both OPEs are more favorable than those reported for other well-known dyes used to detect Aβ aggregates including ThT and CR. While the entire protein structures were different in the referenced studies, the common cross-β core of the aggregates were similar. Consequently, this result is consistent with our experimental finding that the OPEs bind to Aβ fibrils with higher affinity compared to ThT. Future simulations are underway by our group to further explore direct comparisons of previous gold standard sensors with OPEs.

Results from this simulation study thus gave molecular level insights as to how and why OPEs are such effective sensors of Aβ oligomers and protofibrils. Their high affinity binding is facilitated by a combination of hydrophobic effects, favorable electrostatic interactions, and OPE complexation. The simulation results also support previously proposed modes of OPE binding fluorescence enhancement, including OPE J-aggregate formation, planarization of the phenyl moieties of the OPE backbone that increases conjugation length, and unquenching of the ethyl ester end groups. Importantly, these mechanisms are not protein-specific, and likely contribute to the experimentally observed sensing of other disease-related protein assemblies, including tau fibrils in AD and α-synuclein aggregates found in PD. Molecular interactions between these aggregate species and OPEs will be investigated in future studies. The multiple sensing modalities provided by the OPE compounds make them promising agents for the screening of amyloid protein aggregates with more robustness than currently available dyes.

MATERIALS AND METHODS

Initial Structures

The initial coordinates of OPEs and Aβ aggregates were assembled with UCSF Chimera⁶⁴ by placing the OPEs a minimum of 10 Å from the oligomer surfaces. The fibril structure, 2LMN⁶⁵ was obtained from the Protein Data Bank. In this work, we prepared two different sizes of Aβ oligomers from the fibril structure: a 5-mer and a 24-mer. The 5-mer was obtained by saving the coordinates of five consecutive peptide chains from the original PDB structure from the protein databank. In order to build the larger 24-mer, we extended the 2LMN structure one monomer at a time until each fold contained twelve monomers each. This allowed us to increase the β-sheet rich surface area to model a larger fibrillar conformation. The 24-mer was then simulated in absence of OPEs for 100 ns in order to allow the new structure to stabilize. Only single trajectories were obtained for the 24-mer system due to computational constraints. We note that these protein structures are open at the ends of the growth-axis allowing availability for OPE binding. We chose these particular oligomer structures for this study in order to serve as a starting point for simulations of OPE-Aβ that could be compared directly with other ligand binding studies by our group.⁴⁹ The β-sheet rich core is similar to the aggregate structures of Aβ investigated by Nowick et al.⁶⁶ Future studies could be carried out to explore OPE binding to other neurodegenerative-based protein morphologies. OPEs were built using GaussView 5 package.⁶⁷ Geometry optimizations were carried out with Gaussian 09⁶⁷ as previously described.⁶⁸ Table 2 summarizes the four systems used.

Table 2.

Details of simulated systems in this work

System	# of OPEs	Net charge per OPE	[OPE] mM	# of Aβ peptides	# Trajectories
I-A	10	−2	8.5	5	3
I-B	10	+4	8.5	5	3
II-A	12	−2	8.7	24	1
II-B	12	+4	8.7	24	1

Simulations were prepared using the AmberTools suite.⁶⁹ Parameters for simulating the protein structures were obtained from the Amber14 force field.⁷⁰ The generalized Amber Force Field (GAFF)⁷¹ was selected to parameterize the OPEs with partial atomic charges derived using the Restrained Electrostatic Potential (RESP)⁷² approach via the R.E.D. (Restrained Electrostatic potential charge Derive) server.⁷³ Each system was solvated with explicit water molecules (TIP3P model)⁷⁴ and counterions were introduced to maintain charge neutrality.

Molecular Dynamics

The AMBER MD package was used to carry out all simulations.⁷⁵ The process of producing these systems started with 1,000 steps of energy minimization holding the Aβ oligomers and OPEs fixed to allow equilibration of water molecules, followed by 2,500 steps of energy minimization of the entire system. The system was then heated to 298 K, followed by NVT (constant number of molecules, volume, and temperature) equilibration at a constant volume for 0.5 ns. Lastly, full production NPT (constant number of molecules, pressure, and temperature) simulations were performed for 100 ns. Temperature was regulated using Langevin dynamics with a collision frequency of 2 ps⁻¹, and a new random seed was explicitly specified for each simulation restart. A timestep of 2 fs was used. The nonbonded cutoff distance of 8 Å was used for short range effects with use of the Particle Mesh Ewald sum (PME) for long range electrostatics. Computing resources provided by the Extreme Science and Engineering Discovery Environment (XSEDE)⁷⁶ via Comet of the San Diego Supercomputing Center was used for MD simulations with GPU acceleration (with default single precision fixed point or SPFP). After production runs, the trajectories were merged and processed using PTRAJ and CPPTRAJ,⁷⁷ and visualized using UCSF Chimera.⁶⁴ Binding energy values were determined using the molecular mechanics Generalized Born Surface Area (MM-GBSA) method.⁷⁸ The implicit solvent (GB) model used here was developed by Onufriev and colleagues.⁷⁹ Additional details of MM-GBSA as applied here can be found in our previous work.⁶⁸