Destabilization of Human Serum Albumin by Ionic ... - ACS Publications

Subscriber access provided by UB + Fachbibliothek Chemie | (FU-Bibliothekssystem)

Article

Destabilization of Human Serum Albumin by Ionic Liquids Studied Using Enhanced Molecular Dynamics Simulations Vance W. Jaeger, and Jim Pfaendtner J. Phys. Chem. B, Just Accepted Manuscript • DOI: 10.1021/acs.jpcb.6b09410 • Publication Date (Web): 03 Nov 2016 Downloaded from http://pubs.acs.org on November 5, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

The Journal of Physical Chemistry B is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

Destabilization of Human Serum Albumin by Ionic Liquids Studied Using Enhanced Molecular Dynamics Simulations Vance W. Jaeger and Jim Pfaendtner* Department of Chemical Engineering, University of Washington, 105 Benson Hall, Box 351750 Seattle, WA 98195

*Corresponding Author Email: [email protected], Phone: 1-206-616-8128, Fax: 1-206-685-3451

ACS Paragon Plus Environment

1


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 30

Abstract Ionic liquid (IL) containing solvents can change the structure, dynamics, function, and stability of proteins. In order to investigate the mechanisms by which ILs induce structural changes in a large multi-domain protein, we study the interactions of human serum albumin (HSA) with two different ILs, 1-butyl-3-methylimidazolium tetrafluoroborate and choline dihydrogenphosphate. Root mean square deviation and fluctuation calculations indicate that high concentrations of ILs in mixtures with water lead to protein structures that remain close to their crystallographic structures on timescales of hundreds of nanoseconds. To overcome potential timescale limitations due to the high viscosity of the solvent, we employed enhanced sampling techniques to estimate the free energy of an experimentally determined important transition within the protein structure. Metadynamics simulations show that the free energy landscape of the unfolding of loop 1 of domain I is different in the presence of ILs than it is in water, consistent with previously published experimental evidence. We then apply essential dynamics coarse graining to systematically predict differences in the dynamics of proteins solvated in ILwater mixtures versus pure water systems. We also demonstrate that the presence of ionic liquids changes the distribution of intermolecular distances among several ligands, indicating that the protein structure swells in the presence of certain ILs, consistent with experimental evidence.


2

Page 3 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction Ionic liquids (ILs) are a class of solvents that consist of salts that are liquid at temperatures below 100° C. Because of the unique ionic structure of these solvents, they tend to have a set of properties that make them useful in several applications. ILs tend to have low volatility, high thermal stability and low flammability compared to some of the most common organic solvents, leading to their frequent classification as green solvents.1 ILs are useful for the solvation and separation of biomolecules related to biofuel synthesis. For example, lignocellulosic biomass processing can be enhanced by the addition of certain ILs such as [EMIM][OAc] which has the rare ability to solvate crystalline cellulose.2,

3

Algal biomass

processing can be enhanced by using ILs to degrade the microbe’s cell walls, releasing the important lipids within.4 ILs have been shown to be useful in a number of applications related to chemical synthesis and catalysis. For example, IL-containing solvents can enhance reaction rates when combined with transition metal catalysts as well as change enantioselectivity or reaction equilibria when combined with enzymes such as lipases.5, 6 All of the above properties make ILs an interesting solvent for enzymatic hydrolysis of biomass. In addition to characteristics that make ILs suitable solvents for chemical processes and catalysis, ILs can also be used in mixtures that stabilize proteins for long-term storage.7-9 This opens possibilities to extend the shelf life of therapeutic or industrially relevant proteins. In spite of these many applications, widespread use of IL/biomolecular systems in commercial or industrial applications have been stymied due to the occurrence of deleterious changes in protein structure and function – an unintended consequence of significantly changing the native environment from which the protein or enzyme originated. There is still a need to understand at a mechanistic level why some ILs lead to loss of biomolecule functions and others


3


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 30

do not. While there are some heuristic rules based on the Hofmeister series for choosing protein stabilizing ILs, there is a lack of fundamental understanding of what makes a favorable ILprotein combination, whether it is for enzymatic hydrolysis of biomass or for long-term storage of proteins10-13. Several previous studies have demonstrated that IL-induced changes in protein structure can be described or even predicted by molecular dynamics (MD) simulations14. There is a wide range of IL-tolerance displayed by different enzymes, and the stability of an enzyme depends on both the species of the IL and the specific enzyme used in the research. Several enzymes have been shown experimentally to tolerate the presence of ILs in concentrations of up to 20% v/v in water before deactivating; others have been shown to tolerate neat ILs and retain high levels of catalytic activity. The IL-tolerance of several different enzymes, including xylanase, lipase, chymotrypsin, and several cellulases has been studied by MD simulations and with various experimental techniques.15-22 Specific interactions between cations, anions and the proteins of interest were analyzed to determine why certain ILs stabilize or destabilize these enzymes. Typically, the analysis of IL-protein MD simulations includes many standard measures of protein or solvent structure and dynamics including: root mean square deviation and fluctuation (RMSD and RMSF) of protein structure, principle component analysis (PCA), interaction lifetimes between ions and catalytic residues, surface residue entropy, structuring of water and ions around the protein, secondary structure evolution, among others.23 Frequently proteins remain very close to their crystallographic structures during molecular dynamics simulations even when they are solvated in very high concentrations (50% w/w or more) of IL mixed with water. It often takes hundreds of nanoseconds or more to observe significant structural changes beyond thermal fluctuations at these concentrations, even when it is known experimentally that the protein unfolds quite readily in ILs. One possible explanation


4

Page 5 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


for the slow evolution of the protein structure in the presence of high concentrations of ILs is that the high viscosity of the solvent retards the exploration of sidechain and backbone conformations. Herein, we supplement classical MD simulations with free energy calculations using the metadynamics method and a coarse-grained analysis known as essential dynamics coarse graining (EDCG). These methods allow simulations to more effectively sample the conformational changes of a large multi-domain protein in viscous media, and to determine the differences in protein structure in IL-containing solvents versus water. Metadynamics adds a time-dependent potential bias to a collective variable of interest within the protein structure. The bias is added in such a way that the free energy landscape of the collective variable can be uncovered.24,

25

EDCG is a method by which continuous domains of the protein are assigned

based on a minimization of intradomain covariance and maximization of interdomain covariance.26 To date, most simulations of biomolecules in ILs have used classical MD exclusively or docking protocols coupled with spectroscopy27, 28, with the exception of two studies investigating sugar conformational switching in IL29, 30 and a recent study quantifying the rate of mini-protein unfolding in ILs. To test the utility of metadynamics and EDCG for studying protein conformational switching in ILs, we studied human serum albumin (HSA) as a test system. From the point of view of simulations, HSA is ideal given that it has well-defined subdomains from which to benchmark EDCG and that experimental evidence indicates that a specific loop in the protein denatures in the presence of 1-ethyl-3-methylimmidazolium tetrafluoroborate ([BMIM][BF4]).31 In the case of HSA there is a very specific region (loop 1 of domain I) that can be probed in these simulations using metadynamics. Moreover, the same experimental study, upon which we are basing our simulations, indicates that coupling of HSA’s domains’ motions


5


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 30

are affected by [BMIM][BF4]. Covariance analysis, principle component analysis and EDCG can characterize this coupling in simulations. Additionally, the interactions between [BMIM][BF4] and the closely-related bovine serum albumin protein have been studied experimentally using fluorescence spectroscopy, circular dichroism, and docking simulations, which found that the protein swells in the presence of the IL, the enzymatic activity of the protein is moderately reduced by the IL, and that specific interactions of the IL and the protein can be investigated using computational methods.32 Ultimately, bovine serum albumin is unfolded by [BMIM][BF4] at high concentrations. Another interesting study shows that choline dihydrogen phosphate ([chol][dhp]), in mixtures with water, stabilizes the structure of HSA, wherein the protein binds fatty acids similarly in the presence of ILs and in crystallographic structures.33 In this same paper, HSA is found to swell in the presence of [chol][dhp], but it is unknown whether the nature of this swelling is the same as observed in bovine serum albumin in the presence of [BMIM][BF4] . In reality, neat [chol][dhp] is not technically an IL, its melting temperature is 392K which is above the 373K threshold that is widely given as the threshold for the definition of an IL.34 It is instead more properly an organic salt with a high solubility in water. For simplification, in the rest of this discussion, we will include [chol][dhp] as an IL. The imidazolium-based IL, [BMIM][BF4], on the other hand severely hampers the ability of HSA to bind fatty acids. We analyzed the differences in the structures of the binding domains over the length of the simulations to determine the extent to which MD can capture this behavior. Ultimately, the results presented herein indicate that our simulations can be used to (a) accurately replicate known IL-induced structural transitions in large proteins, (b) provide additional molecular-level detail on interactions between ILs and proteins, and (c) suggest which regions of


6

Page 7 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


the protein have the largest structural changes when the solvent is varied. This is of particular interest for large proteins and highly viscous solvents that require long timescales to evolve.

Methods Molecular Dynamics. Two molecular structures of HSA were taken from the protein databank; one of which contains fatty acids bound to the protein while the other does not (PDB: 4IW2, 1E7I).35, 36 Protonation states of sidechains were determined using PROPKA3.037 at the biological pH of 7.4. The protonation states of these sidechains are specified in the supplemental material. All molecular simulations were performed in the GROMACS 4.538 simulation package patched with PLUMED 2.039. The AMBER99SB-ildn40 force field was used to treat protein interactions, water was simulated using the TIP3P model, a combination of GAFF41 and a published force field42 was used to treat [BMIM][BF4], and GAFF was used to treat [chol][dhp] and stearic acid. The electrostatic point charges of the ionic liquid, where derived and not taken from literature, were developed using the HF/6-31G(d) level of theory in Gaussian0943 and the RESP method44 in the antechamber force field development package. Force field parameters developed for these ILs are provided in the supplemental information. Lennard-Jones forces were shifted to zero beyond 1.2 nm. Electrostatics were treated using particle mesh Ewald summations. A time step of 2 fs was used in all simulations, and the LINCS algorithm was used to constrain the bonds between hydrogens and heavy atoms. Pressure was maintained at 1 bar with the Berendsen barostat45, and the temperature was controlled with a stochastic global thermostat46. Data from simulations at 310K and 343K will be presented in the main text. Molecular dynamics simulations began with 10,000 steps of steepest descent minimization followed by 5 ns of solvent equilibration (i.e., frozen protein) at 500K in the NVT ensemble. The


7


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 30

subsequent production simulations were 200 ns, with three total replicas, one for each temperature (293K, 310K, 343K) for each solvent condition. Metadynamics. To study conformational transitions within HSA, we constructed a model system using a portion of the protein as presented in Figure 1A. Metadynamics simulations were conducted using the above parameters with a smaller section of the protein (SER5 to PRO113) at a temperature of 310K. Pressure control was removed to sample the canonical ensemble. This section of the protein was selected because it contains loop 1 of domain I, which has been shown experimentally to denature in [BMIM][BF4], along with nearby structural elements.31 Metadynamics is a method to bias simulations away from previously explored conformations so that a full range of structures and conformations can be explored and thereby provide estimates of the relevant free-energies in the system (as projected onto reduced dimensionality descriptors called collective variables or CVs). Further discussion about metadynamics is provided in the supplemental information.


8

Page 9 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 1: (A) Domain I model used for metadynamics simulations (SER5-PRO113). Black spheres are restrained atoms. Red and cyan spheres represent the atoms used for the centers of mass for the distance CV. (B) Initial structure of domain I (white). Structure after 200 ns of MD. 20% [BMIM] [BF4] (red). 98% [BMIM] [BF4] (magenta). 20% [chol] [dhp] (blue). Water (gray). To study the loop rearrangement, we biased the distance between the center of mass of the Cα atoms of residues 32-35 (group 1) and 80-86 (group 2) in a similar manner to our previous work studying binding pocket opening/closing in actin.47 The initial height of the


9


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 30

Gaussians was 1.0 kJ/mol-nm, which decayed with a bias factor (γ) of 20; the width was 0.03 nm and deposition rate was one hill per 5.0 ps. A harmonic restraint of strength 500 kJ/mol-nm2 was placed on the upper bound of the CV above 1.8 nm. An additional restraint (strength 500 kJ/molnm2) was placed on the RMSD (for values of RMSD > 0.14 nm) of the alpha carbons at important areas of the protein to keep the unbiased portion of the fragment near the crystallographic structure of the full protein. Essential Dynamics Coarse Graining. Essential dynamics coarse graining (EDCG) is a method by which one can define continuous domains of a protein that can act as coarse-grained sites.26 This method was applied to HSA to understand if its slow modes of motion are similar in water and in high concentrations of IL. Since RMSD and RMSF are not usually adequate to provide information about long timescale transitions in an IL-solvated protein’s structure, EDCG was tested to determine if it can predict sites that are most affected by the new solvent environment so that later simulations or experiments on those specific areas of the protein’s structure. By analyzing MD trajectories of the protein in water and in IL-water mixtures, covariance matrices are constructed to quantify the correlation of the atoms’ motions. The covariance of atom i and atom j in dimension h and k is defined in Equation 1, where C is the magnitude of covariance, n is the number of frames from the trajectory, t is time, and ∆r(t) is the difference between the current position of the atom and the average position of the atom over all frames of the trajectory. This leads to a matrix of values that is of the size 3N x 3N for a system of N particles. In practice, only alpha carbons in the protein backbone are used. (Eq. 1)

( , ) ≡ ∑ , () , ()

The covariance matrix is diagonalized to calculate eigenvectors that are known as principle component analysis (PCA) modes. The modes are ordered by the magnitude of their


10

Page 11 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


eigenvalues to understand which modes contribute the most to long timescale structural transitions. This method allows us to ignore those motions that are fast and localized and do not contribute to large conformational changes. Next, the number of CG model sites is selected. After trying 6, 12, 18 and 24 sites, an 18site model was settled upon. This produced domains that ranged in size from several residues to tens of residues. Domains of this size have CV-descriptors that can be biased with metadynamics. The alpha carbons included in each CG site are chosen by an algorithm that minimizes, by steepest descent, the intrasite covariance while maximizing the intersite covariance. There are three constraints to the selection of sites. First, each atom is allowed to be in only one site. Second, the CG site is located at the center of mass of the alpha carbons in the group. Third, the atoms are assumed to be part of a continuous protein sequence.

Results and Discussion Structural Evolution. The initial and final structures of the 200 ns ligand-free MD simulations are illustrated in Figure 1B. As a general measure of the stability of ligand-free HSA in each solvent, the root mean square deviation (RMSD) of the alpha carbons from the crystallographic structure was measured over the length of the MD simulations. Additionally, the root mean square fluctuation (RMSF) of each alpha carbon from the average simulation structure was calculated. These data are presented in Figures 2 and 3 respectively for each concentration of solvent at 310 K. Data at the two other temperature replicas (293K, 343K) are provided in the supplemental information as Figures S2 and S3, and similar trends are observed for these replicas. RMSD and RMSF tend to increase with increasing temperature, as expected. Most interestingly, we observe that the RMSD of HSA solvated in [BMIM][BF4] at 343K increases to magnitudes observed in water-solvated HSA simulations at 310K. This supports the hypothesis


11


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 30

that the relatively low RMSD observed for [BMIM][BF4] may be due to the high viscosity of the solvent (which is reduced at higher temperatures) coupled with timescales that are too short to observe large structural transitions rather than a true stabilization of the protein in the high concentration IL solvent. For RMSF graphs, we truncated the presentation at residue 530 because of some instability near the C-terminus that causes large fluctuations. The RMSD data demonstrate that at these timescales, HSA remains closer to its crystallographic structure in high concentrations of IL than it does in water. This is a symptom of the limited timescales we can explore with MD. In 98% [BMIM][BF4] HSA remains trapped near its crystallographic structure. The RMSDs of these IL systems demonstrate why we need enhanced sampling techniques to gain insight into structural transitions in these highly viscous media. As for the systems containing 20% IL, we observe that RMSD is similar to that of water. This is consistent with simulation results in our previous work which that low (

Destabilization of Human Serum Albumin by Ionic ... - ACS Publications

Recommend Documents