Effect of the electrostatic surface potential on the oligomerization of full-length human recombinant prion protein at single-molecule level

The electrostatic potentials were known to have essential roles in the functions and physiological activities of proteins.^1–4 Many fundamental biological processes can be regulated by the electrostatic interactions among biomolecules such as proteins and nucleic acids.^3,5–7 In many cases, the major electrostatic interactions during the protein reactions are determined by the electrostatic surface potentials (ESPs) of the active sites located on protein surfaces because most charged amino acid (AA) residues tend to locate on the protein surface.^8–10 Additionally, the surface charged residues help to stabilize the conformations of proteins or protein aggregations.^10–13 The most widely used model for ESP calculation of proteins is the Poisson-Boltzmann equation, with the assumption that the solvent is a high-dielectric continuum and the protein is a low-dielectric cavity.¹⁴ The ESP of protein can change with the solution pH, because the charges carried by AA side chains could have major influence on the folding structure of the protein, and in turn determine the ESP distribution.^14–16 For example, the α-helix and β-sheet are the most common secondary structures in protein folding conformations, and the major contributions of their interactions with other domains or other proteins are hydrogen bonds and electrostatic interactions, which are directly connected to the charges of AA residues in those domains.^17,18 Therefore, both intramolecular and intermolecular interactions during the protein folding and aggregation can be affected by the change of ESP distribution on the protein surface.^19,20 ESP distribution can thus be used as an important factor to control the folding conformations and behaviors of biomolecules, such as their activities under various physiological conditions.¹⁴ One of the important topics in this field is the relationship between the ESP distribution and protein aggregations involved in the amyloid diseases.²¹ Many experimental and theoretical studies have shown that the conformational changes of intrinsically disordered proteins (IDPs), such as the prion protein (PrP), were induced by the changes of environmental pH values, which directly affect the ESP distributions of different domains inside the molecule and the interfaces among different molecules.^21–26,17 To unravel the process of the early stages of PrP aggregations, the molecular basis of the ESP distribution determined by the environmental pH values is critical.^17,27 The pH values of the solutions have been found to directly influence certain protein domains such as α-helices and β-sheets, so that the entire conformations of the protein will follow the change of the ESP and show different activities. In order to have an in-depth understanding of the ESP and its effect to single molecular PrP aggregation process, the only way is to combine the experimental and theoretical tools.

It has been widely accepted that the misfolding of PrP and its aggregation is directly related to its neurotoxicity in the brain tissues.^28–30 The pH value and the charged membrane surfaces have been demonstrated to induce the misfolding of PrP, and in turn result in large amyloidal structures.^31,32 However, the initial physical and chemical mechanisms of this aggregation process remain unknown, and the ESP distribution and the correspondent structural changes have not been investigated yet.^31,33 Clearly, the conformational changes of PrP molecules and their ESP distributions in different pH environments are connected to the aggregation mechanism.³³ Various techniques have been used to study the mechanism of large amyloidal aggregates of PrP.^34,35 However, the initial formations of small PrP oligomers, such as dimers and trimers at nanometer scale, deserve more attention and careful studies, which may foster the prevention and treatment of prion related diseases.^36,37 These small oligomers act as the basic binding units to the subsequent aggregations of large amyloidal fibers.^38,39 The structures and ESP distributions of these small oligomers are difficult to determine, due to their intrinsically unstable misfolded structures.^40,41 Most previous studies on the possible effect of ESP to the PrP oligomers were based on vague schematics, which require the direct experimental and theoretical scrutinities.^42,19

Atomic force microscopy (AFM) has been successfully used to obtain surface morphology information of PrP peptide aggregations at single-molecule level.^19,43 Therefore, it is feasible to explore the role of the ESP in the mechanism of full-length human recombinant prion protein (sequence: 23-231) oligomer aggregation via the AFM images of the PrP oligomers, which have their corresponding ESP distributions and may form specific patterns on different substrates, and under different pH conditions.^38,44 Here, the mica surface was used to mimic the hydrophilic surface of cell membrane. The PrP monomer (M) folding conformations at different pH values were predicted by molecular dynamics (MD); the structures of the dimer (D), and trimer (T) were obtained from molecular docking simulation using the M unit;⁴⁵ and the conformations of larger oligomers such as M-D and D-T were predicted with the help of high-resolution AFM images, by the same method used in previous publications.^38,44 Based on these structures simulated by MD and docking, the ESP simulations were conducted, and the detailed aggregation mechanisms at single-molecule level have been established.⁴⁶ The combination of the surface morphology measurements and structural simulations successfully revealed the molecular basis of the ESP distributions involved in PrP oligomerization.

Frozen full-length human recombinant prion protein (sequence: 23-231) was purchased from Calbiochem® in Germany with a concentration of 2 mg/ml, which was from E. coli expression and purified as previously described.¹⁰ The purity of >95% was determined by SDS-PAGE. The protein was stored at −20 °C in 10 mM sodium acetate buffer at pH 4.0 before usage. Before each experiment, the prion protein solution was centrifuged (20 000 g) for 30 min at 4 °C to remove pre-existing aggregates. Phosphate buffered saline (PBS: 100 mM sodium phosphate, 150 mM sodium chloride, pH 7.2, added with 0.05% sodium azide) was purchased from Pierce (Themo Scientific, Waltham, MA, USA). Triple deionized water was provided by Barnstead Nanopure Diamond Laboratory Water System. Mica was purchased from Ted Pella, Inc. (Product No: 56).

An Agilent 5500 Controller combined with a 50 μm × 50 μm Agilent multipurpose AFM scanner (S/N: 325-00388, with 0.5 Å RMS of z noise specification) was used to obtain images in an area as large as 1000 nm by 1000 nm. All the images were taken under TopMAC mode with a PicoTREC controller (Agilent Technologies, Santa Clara, CA), which is based on a single, low-force touch of the cantilever tip, enabling the non-intrusive imaging of PrP aggregates on the surfaces. The whole system was shielded from environmental interference by a PicoPlus Isolation Chamber. Silicon cantilever tips with spring constant of around 0.1 N/m were used for experiments.

The detailed parameters for the image are as follows: drive is approximately 20%, resonance gain is 2, resonance frequency is around 6.725 KHz, resonance amplitude is about 4.837 V, and scan rate is about 1 line/s.

PrP oligomers were formed at the concentration of 100 ng/ml in pH 4.5 acetate acid buffer, and in pH 7.2 PBS buffer as contrast. And 400 μl of each sample was deposited onto freshly caved mica surface in liquid cell for AFM scanning immediately at room temperature at 20 min intervals. All images were acquired by the TopMAC mode, which enables non-intrusive imaging of weakly attached PrP aggregates with both lateral and vertical subnanometer resolution. All of the AFM imaging was done in liquid with the AFM tip immersed in the solution. The AFM images were obtained with 1024 × 1024 points.

The AFM images were processed using WSxM software.¹¹ If necessary, only first-order plane-fittings were performed with the analytical software. The height values of the PrP aggregates were determined by individual cross section profiles generated by the software. To ensure correct height data, the height was measured twice, at both the cis- and trans-positions, and averaged.

Currently, Poisson-Boltzmann equation^5,9 is one of the most used theoretical models for electrostatic interactions calculations. In a simple system of charged particles in homogeneous medium (solvent), the electrostatic potential ψ(r) can be described with the space-dependent dielectric constant ε(r) and the charge density ρ(r) of the solute.⁹ According to this approach, the PrP oligomer (M, D, or T) in solution can be treated as particles with an interior dielectric constant (ε_int = 4), and the solvent can be considered as a continuous medium with an exterior dielectric constant (ε_ext) of 80.⁹ Three assumptions are important in order to apply Poisson equation to the system of PrP oligomers. The first assumption is that the gradient of the electrostatic potential ψ at any surface position r corresponds to the “equivalent force” applied to that particle coordinate by the whole system, including both solvent ions and the solute molecules.⁹ The second assumption is that, in this solvent-solute system, the potential generated by the ionic distribution in the solvent equals to the electrostatic potential at the surface position r multiplied by the charge value of solvent ions at that position.⁹ The third assumption is that the ions in the solvent follow Boltzmann distribution, so that the cause of the total electrostatic potential is the linear combination of the charge distribution in the protein and the Boltzmann distribution of the ions in the solvent.⁹ 

In the special case of PrP oligomerization, the ψ(r) value at some surface positions may not be small enough for the linear approximation, and the geometry of the surface also influences the local potential.⁹ Therefore, the detailed investigation on the distribution of PrP oligomer ESP and its possible change during the oligomerization reaction requires the help from the structural information of those PrP oligomers. In this study, the stable 3-D structures of PrP oligomers were predicted from MD and docking simulations. Based on the simulated structures of PrP oligomers, the numerical solutions of their ESP values were obtained by DelPhi web server.^46,47

In DelPhi simulations, the surface of the PrP oligomer is divided into the lattice of grid points with grid resolution of 1 grid/Å by a spherical solvent probe, similar to the scanning of molecular surface with the AFM tip.⁴⁸ After the construction of this solvent-accessible surface (SAS) from the van der Waals radii of the atoms, the induced grid charges on the surface were determined by finite-difference formalism.⁴⁸ In general, this numerical method is based on the grid charge at each high-resolution grid point.⁴⁸ However, using the grid-independent procedure to calculate the potential caused by the real charges on molecular surface, both linear and nonlinear Poisson-Boltzmann equations can be solved with improved precision.⁴⁹ For this study, some parameters used in the DelPhi web server include the following: the atom charges and atom radii are chosen to be based on the Amber force field; the dielectric constant for the protein and the solvent are 4 and 80, respectively; and the non-linear solver is used. All other parameters are set to their default values.

In our study, the single molecule imaging method is not designed to measure the solvation energy directly. However, the changing patterns of the stabilized prion oligomers on the mica surface under different pH conditions indicated that the prion molecules aggregated using different conformations, whose solution free energy levels have been stabilized to different values after the aggregation process. Therefore, the aggregation patterns under different pH conditions are directly related to the ESP distributions of PrP oligomers and also reflect their different stable energy states.^16,50 The AFM images provide useful information not only for the single molecule ESP study but also for the thermodynamic and kinetic studies of the PrP aggregation process.^38,44

The initial protein data bank (PDB) file of PrP M at neutral pH was modified from PDB entry 2LSB. All histidine residues (pKa1 = 2.3, pKa2 = 6.0, and pKa3 = 9.6) in the molecule were set to be positively charged for the later simulation at pH 4.5. The structure was first minimized in vacuum and then put into a solvent box of TIP3 water for explicit solvent simulation. This simulated PrP folding structure in solvent box was an isobaric-isothermal ensemble (NTP), with 1 atmosphere and 300 K as final pressure and temperature. The MD was conducted for 40 ns before the folding structure became stable conformation. The longer time (60 ns) was also used for the simulation and was found not to have noticeable changes of the folding structure. All monomer folding simulations were conducted with Amber 11 molecular dynamics package.⁵¹ 

The PrP M folding structure at pH 4.5 was used for two-body docking to generate D structure, and three-body docking to generate T structure in HADDOCK webserver.⁴⁵ The residues in the two β-sheet regions of the Ms were set as predicted active residues, and all histidine residues were set to be positively charged. For each docking simulation, the HADDOCK webserver generated a series of docking structures, and they were ranked according to the energy levels. The most stable structures were selected as PrP D and T, respectively. The large aggregates formed by M, D, and T were simulated by the same method, but M, D, and T were used as the basic docking units to build the larger oligomers. After the stable M, D, and T structures were obtained from MD and docking simulations, their atomic coordinates were uploaded to the DelPhi web server and used to define the molecular surfaces, where their corresponding ESP distributions were calculated and visualized. Therefore, the MD and docking simulations generated the structural basis for the ESP calculations, and then the ESP calculations highlighted the mechanism behind the PrP folding and aggregating processes. The simulated structures of M, D, and T simulated by MD are shown in Fig. 1 which is used to illustrate the PrP aggregation model based on Poisson-Boltzmann equation.

The single molecule PrP aggregations were revealed by high-resolution AFM imaging. In the representative AFM images shown in Fig. 2(a), the large oligomers formed in pH 4.5 buffer demonstrate clear specific patterns formed by smaller units, M, D, and T. However, the large oligomers formed in pH 7.0 buffer show simple round or rod-like dots, with diameters varying from 10 to 30 nm. These oligomers do not have specific patterns of distinguishable smaller units in their aggregated structures. This difference between the aggregated patterns of PrP oligomers at neutral and acidic pH conditions indicates different aggregation behaviors. Based on previous docking simulations, the binding units of PrP oligomers are M, D, and T in both neutral and acidic buffers, but their ESP distributions are different due to the pH change.³⁸ Therefore, the interactions or aggregating behaviors of M, D, and T were affected by the ESP distributions under different conditions. The AFM images under pH 4.5 imply that the ESP distribution of each PrP unit has greater influence to the conformation of the large oligomers. In other words, the ESP value at the surface of M, D, or T in pH 4.5 buffer can determine the final confirmation of their large oligomer. Moreover, in the acidic environment, the mica surface is negatively charged, so the positively charged surface of each oligomer will have stronger interactions with the substrate.⁴⁴ This interaction is another contribution to the specific patterns shown in AFM images. On the other hand, the hydrogen bonds depend on the AA residues on oligomer surfaces, and most of those residues were not be affected by the pH changes.³⁸ Therefore, the pH-dependence of the ESP on the aggregating conformations also determined the possible change of the hydrogen bonds. The overall effect that caused the difference in the surface patterns of the oligomerization at pH 4.5 and pH 7.0 can be studied together with one mechanism. The AFM images at pH 7.0 only show large amorphous aggregates (Fig. 2(b), numbered as 1′), which have been revealed by previous study to be the self-assembly patterns of M, D, and T.^38,44 The difference from the oligomers at pH 4.5 is that the ESP distributions of these binding units in the aggregation process have less influence on the orientations and conformations of the final aggregated oligomers at pH 7.0. This pH-dependence of oligomer aggregation is clearly related to the electrostatic interactions and ESP distributions revealed form the simulated PrP structures.

From the theoretical perspective, the major interactions involved in prion M, D, and T interactions are hydrogen bonds and electrostatic interactions. Based on our previous study, the stable D and T units were quickly formed via the β-sheet regions of the M units in the solution, then these M, D, and T approached the mica surface and continued to aggregate with each other on the surface to form larger oligomers.^38,42 The binding regions for the aggregations among M, D, and T have been found to be their hydrophilic α-helices regions since the β-sheets of D and T were already occupied to form their hydrophobic core.⁴⁴ Moreover, the mica surface is hydrophilic and has stronger interactions with the α-helices of D and T units.⁴⁴ When the aggregating happened among the binding units to form large oligomers, the dominant interaction to determine the oligomer patterns is the hydrogen bond, because of its greatest bond energy among all non-covalent interactions. However, the electrostatic interaction is much more sensitive to the change of buffer pH, so it also plays critical roles in the PrP aggregation process.^52,53

In order to find the main factors (such as ESP and hydrogen bonds) that determine the binding orientations and conformations of PrP oligomers at different pH values, the ESP distributions of the binding units M, D, and T and larger oligomers were first studied with structural simulations. Fig. 3 shows the simulated ESP distribution of prion M, D, and T at pH 4.5 and pH 7.0, respectively. The ESP for M clearly showed the difference of ESP distributions at different pH values, and it also determined the ESP distributions of D and T, respectively, at pH 7.0 and pH 4.5. The major contribution to the charges of prion oligomers at pH 4.5 is the protonated histidine residues. The charge distributions influenced the folding structure of M, and in turn lead to different binding structures of D and T. The ESP distribution of the M unit clearly becomes more polarized under acidic condition (Fig. 3(b)). All of these units become more positive at pH 4.5, while the D and T units show more dramatic changes of their conformations and ESP distributions, which is clearly illustrated by the change of ESP distribution of M under different pH values. The folding structures of M, D, and T were obtained from MD simulations to define the stable protein surfaces, and then the ESP simulations were conducted based on these surface structures. In MD simulations, the histidine residues were set to be protonated, become their side chains, have the major changes in their structures, and charges when pH decreases from 7.0 to 4.5. Other residues such as aspartic acid and glutamic acid may also have some of their side chains partially protonated at pH 4.5, but it is impossible to accurately predict the protonation state for each residue. The DelPhiPKa web server was used to predict the pKa values for all charged residues in the M, D, and T structures.^54,55 The results confirm that histidine side chains are the major factor in the PrP folding. In the later ESP simulations, all of the residues are included through the simulation process, so the ESP distributions shown in Fig. 3 represent the electrostatic properties of the entire structures.

The ESP values for those individual surface positions of the M, D, and T in Fig. 3 were extracted and compared in the histograms in Fig. 4. The individual ESP value ranges from +500 kT/e to −500 kT/e, but most of them are in the range from −1.5 kT/e to 1.5 kT/e, as shown in the histograms. In Fig. 4(b), the histogram of M at pH 4.5 clearly shows more positive values (highlighted in blue), mainly because of the protonation of histidine residues. On the contrary, the histogram of M at pH 7.0 is more symmetric (Fig. 4(a)), showing the less charged surface compared to the one at pH 4.5, and the more even distribution of ESP at neutral pH. All M, D, and T have the similar trend, with their ESP distributions shifted to the positive direction (more blue areas) under pH 4.5. These histograms directly show how the PrP M, D, and T units are polarized in the acidic pH solution.

The further comparison of the PrP oligomers is focused on the contact interfaces of M, D, and T in their corresponding large oligomers. Based on the AFM images and ESP simulations, the binding regions on those interfaces that determined the conformations of large structures have clearly shown their pH-dependency. Some enlarged AFM images of those basic binding units M, D, T, and larger oligomer structures in pH 4.5 are shown in Fig. 5, and their various conformational morphologies can be illustrated by the structures predicted by docking simulations. The AFM images show relatively loose contacts at the interfaces of M, D, and T. Therefore, the binding residues for M, D, and T were predicted to come from the regions of α-helices, which locate in the outer parts of these binding units. The docking simulations of large oligomers are highly consistent with the experimental results, and each M, D, or T in the predicted structures can be matched with the detailed features in their corresponding AFM images.

The mechanism of aggregation process at pH 4.5 is illustrated in Fig. 6. The M-D-D oligomer in the AFM image has two interfaces, one for M-D and the other for D-D. The M-D interface is used as an example. At pH 4.5, the M shows a dipole moment in its ESP distribution, and the D has two internal dipoles moments. The binding process to form the M-D structure caused the redistribution of the ESP for both of them. However, the new internal dipole moments in M-D structure did not change significantly. The other surface areas of PrP binding units maintain their ESP distributions relatively unchanged, mainly because the relatively large size of PrP molecule compared to the Debye length of ESP. The active interfaces have very little influence on other parts of the oligomer surface. Overall, the dipole moments of M and D helped their approaching and binding process to form the larger M-D oligomer. The same mechanism is also valid for M-T and D-T interface interactions.

In order to further confirm the mechanism illustrated in Fig. 6, the height values of the basic units in the larger PrP oligomers were measured in the cross section profiles of the AFM topography images obtained under pH 4.5. The topography images of PrP oligomers, such as the representatives in Fig. 5, were analyzed, and each basic unit was identified according to its different height value. This method has been used in previous publications to study the basic units of PrP large oligomers, and the histogram shown in Fig. 7 did confirm again that the height values from AFM topography images can be used to distinguish the units M, D, and T in the PrP large oligomers on solid surfaces.^38,44 Therefore, both experimental measurements shown in Fig. 7 and the simulated structures shown in Fig. 5 support the mechanism proposed in Fig. 6 that the aggregation of PrP oligomers is driven by the interactions among individual unit M, D, or T.

Figure 7 is the histogram of the height values of more than 700 M, D, and T units that show their clear difference during the AFM measurements and were then used to distinguish each M, D, and T in the large oligomers. The counted numbers for each unit are compared in the histogram in Fig. 7. According to the area proportion of each peak (Table I), the proportions of the three binding units are 17.7% (M), 46.2% (D), and 36.1% (T), indicating that the D unit has the most numbers, while M has the least numbers. According to the ESP simulations in Fig. 3(d), the unit D shows the distribution of positive ESP at the center of its structure, and negative ESP at the two sides. Therefore, the ESP distribution stabilizes the overall oligomer structure. The unit T has the similar ESP distribution in Fig. 3(f), but the observed number of T is smaller than D in the histogram in Fig. 7 (shown in Table I, the peak area under the green curve is smaller than the one under the blue curve). One reason is that the probability for three Ms to form the T unit is smaller than the one for two Ms to form the D unit. The ESP distribution of M is very asymmetric, which increases the probability of the unit M to interact with each other and form units D and T.

In order to have a better understanding of the contributions of each binding residue in M, D, and T units, the AFM images in Fig. 2(a) were also used for the statistical analysis of the predicted binding residues at the interfaces. Some representative oligomer structures are shown in Fig. 8 for the analysis of residues involved in the PrP large oligomer aggregations. The charged AA side chains are listed in Table II. Most of these side chains locate in the α-helices structures. They helped the individual M, D, or T approach to each other in certain orientation and stabilized the final binding conformations of those large oligomers. The larger oligomer D-D has more charged AA side chains involved at the binding interface. This higher proportion implies that the charged side chains have greater influence on the ESP distribution of larger oligomers such as D-D than the smaller ones such as M-D and M-T.

Figure 8 also shows that the ESP distributions of M-D, M-T, and D-D oligomers all have the positive parts (in blue) in their central core regions, and the more neutral and negative parts tend to point outside. The ESP distribution of D-D oligomer in Fig. 8(c) does show more polarized distribution, with more negatively charged surface areas in the center part and more positively charged surface areas in the outer part. On the other hand, M-D and M-T show relatively mixed distributions of positively charged and negatively charged surface areas in their simulated ESP images. Compared with Fig. 3, it is clear that the ESP distribution of each unit M, D, or T changed during the aggregation process, and this change of ESP facilitated the entire aggregation process in the acidic pH solution, with the help from other types of interactions such as hydrogen bonding.

In each oligomer, the active residues at the binding interfaces for the M, D, and T units have multiple hydrogen-bonds, which are the major driving force for the re-distribution of the ESP during the aggregations (Figs. 8(d)–8(f)). Table III lists the AA side chains predicted in hydrogen bonds among all AA residues involved at the binding interface of PrP oligomers under pH 4.5. For each M-D, M-T, or D-D oligomer, the proportion of AA residues for hydrogen bonds is higher than the one for charged residues. Therefore, the major interaction to determine the aggregation process is the hydrogen bonds, and the ESP plays the additional supporting role. However, the charges on the AA residues at the interface of binding units M, D, and T determine the ESP distribution of that oligomer, as well as how the hydrogen bonds formed through the side chains of those residues. The side chains affected by the pH changes played critical roles in both hydrogen bonds and electrostatic interactions, as Tables II and III showed. Therefore, the overall effect of the hydrogen-bonds and ESP is the stabilization of the PrP oligomers at pH 4.5, and those oligomers may continue to form larger fibrils with their specific conformations. However, the interactions generated by these residues were greatly reduced or altered at neutral pH, so the large oligomers from PrP units M, D, and T became more homogeneous, round-shaped aggregates on mica surface.

In the aggregation processes of PrP oligomers, the ESP distributions of the PrP binding units M, D, and T were directly affected by the pH values. The different ESP distributions further determined the binding conformations of large PrP oligomers on mica surface. The AFM high resolution images were used to monitor the detailed patterns formed by PrP oligomers under different pH conditions. The experimental results agreed with well the theoretical simulations. The combination of the molecular docking and ESP simulations revealed the detailed mechanism of how the PrP oligomers aggregated together through the α-helices domains on the PrP surface. Both experimental and simulation results were used to build the reaction model based on the ESP distributions of binding unit M, D, and T. The ESP distribution at pH 4.5 facilitated the formations of larger oligomers. On the contrary, the electrostatic interactions were weakened at pH 7, and therefore the larger oligomers formed very different patterns. The AA residues involved in hydrogen bonds and electrostatic interactions were also predicted with the help of experiments and simulations. This work provides important information to understand the molecular basis of the protein aggregation on solid substrates.

We thank the National Science Foundation (Nos. ECCS 1231967 and CBET 1139057) for financial support of this research.