TY - JOUR
AU - Otzen, Daniel, E
AB - Abstract Anionic surfactants denature proteins at low millimolar concentrations, yet little is known about the underlying molecular mechanisms. Here, we undertake 1-μs-long atomistic molecular dynamics simulations of the denaturation of acyl coenzyme A binding protein (ACBP) and compare our results with previously published and new experimental data. Since increasing surfactant chain length is known to lead to more rapid denaturation, we studied denaturation using both the medium-length alkyl chain surfactant sodium dodecyl sulfate (SDS) and the long alkyl chain surfactant sodium hexadecyl sulfate (SHS). In silico denaturation on the microsecond timescale was not achieved using preformed surfactant micelles but required ACBP to be exposed to monomeric surfactant molecules. Micellar self-assembly occurred together with protein denaturation. To validate our analyses, we calculated small-angle X-ray scattering spectra of snapshots from the simulations. These agreed well with experimental equilibrium spectra recorded on ACBP-SDS mixtures with similar compositions. Protein denaturation occurs through the binding of partial micelles to multiple preferred binding sites followed by the accretion of surfactant monomers until these partial micelles merge to form a mature micelle and the protein chain is left disordered on the surface of the micelle. While the two surfactants attack in a similar fashion, SHS’s longer alkyl chain leads to a more efficient denaturation through the formation of larger clusters that attack ACBP, a more rapid drop in native contacts, a greater expansion in size, as well as a more thorough rearrangement of hydrogen bonds and disruption of helices. Introduction Surfactant-protein complexes have been studied for decades, thanks not least to their many applications in fields ranging from medicine, detergency, cosmetics, foods, to the oil industry (Aehle, 2007; Otzen, 2011). Critical to these applications is the ability of ionic surfactants to denature proteins at low millimolar concentrations, i.e. 3–4 orders of magnitude lower than the concentrations needed by chemical denaturants such as guanidinium chloride (GdmCl) and urea (Andersen and Otzen, 2009; Otzen, 2015). Anionic surfactants, such as sodium dodecyl sulfate (SDS), (Otzen, 2011; Otzen and Oliveberg, 2002a) are particularly efficient, while cationic surfactants do not always denature proteins (Otzen, 2002). Although complete replacement of ionized side chains does not prevent efficient SDS-induced protein denaturation in practice (Højgaard et al., 2018), protein-SDS interactions are generally thought to be driven both by electrostatic and hydrophobic factors (Jones, 1996), allowing SDS to bind tightly to cationic side chains through electrostatic interactions and penetrating the hydrophobic protein core using its alkyl chain (Yonath et al., 1977). As a result, the surfactant can disrupt both the tertiary and secondary structures (SSs) of the protein, although micellar concentrations of SDS in many cases can preserve and even induce an α-helical SS (Andersen et al., 2009; Mattice et al., 1976; Montserret et al., 2000; Wu et al., 1981). When present in excess, SDS tends to bind proteins at a fixed weight–weight ratio, reflecting complete denaturation and solubilization of the protein polypeptide chain in complex with micelles (Pitt-Rivers and Impiombato, 1968; Tanford, 1980). However, the mechanism of binding of surfactants and subsequent protein denaturation is strongly influenced by surfactant concentration, particularly in the range below the critical micelle concentration (cmc). Indeed, many proteins are unfolded in this range (Otzen, 2011; Otzen, 2015), indicating that monomers as well as micelles can drive surfactant-mediated denaturation. Monomer-driven denaturation has been suggested to involve the formation of small surfactant clusters on the protein surface, in some cases promoted by the association of multiple protein molecules to stabilize a shared micelle (Andersen et al., 2009). The most detailed experimental approaches to elucidate the mechanisms of surfactant-driven denaturation have used the libraries of single-point mutants which remove individual side chain interactions. The consequences of these truncations have been measured by detailed kinetic studies and the results interpreted in terms of different unfolding schemes (Andersen and Otzen, 2009; Otzen and Oliveberg, 2002a). However, such schemes typically assume that the removal of individual side chains will not fundamentally alter the unfolding pathway. While this is generally the case for unspecific denaturant-driven unfolding of proteins, individual side chains can play a major role in protein-surfactant interactions (Otzen et al., 1999). Furthermore, the most detailed study so far on SDS-induced unfolding, using the ribosomal protein S6 as a model system, proposed a model in which proper unfolding was preceded by the formation of a quasi-native S6:SDS complex whose existence could only be surmised indirectly from kinetic data (Otzen and Oliveberg, 2002a) and which implied a fundamental difference from the unfolding route in chemical denaturants (Otzen and Oliveberg, 2002b). In contrast, unfolding of the model protein acyl coenzyme A binding protein (ACBP) in SDS broadly followed the same trajectory as in GdmCl, as seen by the good correlation between unfolding rates of different mutants in SDS and GdmCl, though three outlier mutants were interpreted as a possible early attack around the ligand binding site (Andersen and Otzen, 2009). This makes good sense, given that the binding site is dominated by five Lys residues (13, 18, 32, 50, and 54) which combine to form a highly cationic environment. The bound ligand palmitoyl coenzyme A has features in common with an anionic surfactant micelle. It forms electrostatic interactions with i.a. Lys32 and Lys54 through adenosine phosphate, the pyrophosphate phosphate interacts with Lys13 and the hydrophobic segments of the ligand sequester themselves through intramolecular interactions (Kragelund et al., 1993). Protein engineering analyses yield residue-level information about the structure of the protein during unfolding but do not provide direct information about the way in which, e.g. SDS participates in the unfolding process and the different transient SDS species involved. Small-angle X-ray scattering (SAXS) can, in principle, provide information about the shape of protein-surfactant complexes at different stages of the surfactant micelle aggregation process, as heralded by a recent stopped-flow study of changes in micelle structure induced by salt (Jensen et al., 2014). However, the co-existence of different protein-surfactant complexes makes it unlikely that we will achieve molecular-level insights in the near future (if ever) by such techniques. The experimental timescale of protein denaturation ranges from micro to milliseconds, which was, until recently, longer than timescales readily achievable by direct molecular dynamics (MD) simulation of atomistic protein models. However, steady growth in computational capacity, driven by the appearance of parallel computing platforms, has allowed for more biologically relevant applications of in silico approaches and the gradual convergence between computational and experimental studies in protein folding studies (Rizzuti and Daggett, 2013). Thus, MD simulations have been utilized in studies of proteins’ dynamical features, conformational properties, ligand binding mechanisms, and other properties (Lu et al., 2007; Scheraga et al., 2007; Veitshans et al., 1997; Walker et al., 2013). MD simulations of surfactant-induced denaturation remain computationally challenging because surfactants add another layer of complexity to an already demanding simulation. Surfactants are highly dynamic entities, which can co-exist as monomers and micelles above the cmc and can bind to proteins in a variety of different ways. On the other hand, this complexity makes it difficult to interpret experimental data in isolation and suggests that robust MD simulations could yield otherwise inaccessible insight into the molecular steps involved in surfactant-driven protein denaturation. So far, very few MD studies have probed the mechanisms of surfactant-induced denaturation of proteins (Krishnamani and Lanyi, 2012; Lu et al., 2007; Walker et al., 2013). As early as 2007, Wu and co-workers used a simplified model of a surfactant molecule, where one bead represented the hydrophilic head group and the other bead the hydrophobic tail, whose hydrophobicity could be varied, and analyzed how such molecules interfered with folding of a simplified β-sheet protein in an off-lattice simulation (Lu et al., 2007). In this case, the protein model was a simplified Go-type model in which only native-like contacts are stabilizing. This work did not shed light on how surfactants actively unfold proteins, but nevertheless it showed that the surfactants of moderate hydrophobicity could promote folding if they could dissociate after the initial formation of the collapsed hydrophobic core of the protein. More recently, Sutherland and co-workers (Walker et al., 2013) showed by atomistic MD simulations over a 10-ns time window how SDS micelles were able to support helical structure in an all-helical silk protein. Starting with small disordered micelles binding via hydrophobic, electrostatic, and polar contacts to the protein, the SDS aggregates grew to larger micelles of 40–45 SDS molecules which sustained individual α-helices partially buried within the micelle, though a more detailed quantitative analysis of the degree of interactions between SDS and the protein was not provided. An MD study of helices of bacteriorhodopsin inserted into SDS micelles showed that the four C-terminal helices partitioned to the surface of the micelle to facilitate contacts between charged groups in the transmembrane segment of the helices and the sulfate head groups, though it was concluded that the sampling time may have been insufficient to describe the full unfolding trajectory (Krishnamani and Lanyi, 2012). Self-assembly of surfactant micelles has also been investigated computationally and has to be taken into consideration, since it occurs in parallel with surfactant-protein interactions in computer simulations. In pure water, the aggregation number of SDS micelles increases with SDS concentration and has been experimentally determined to be in the range of 30–80 (Andersen et al., 2009; Bales and Algren, 1995; Bergström and Pedersen, 1999; Berr and Jones, 1988; Chen et al., 1986; Shelley and Shelley, 2000; Vass, 1991). All-atom and coarse-grained MD simulations have revealed the dynamic nature and stepwise growth mechanisms of pure ionic micellar and premicellar aggregates (Pires et al., 2012; Poghosyan et al., 2014). Starting from randomly distributed initial configurations, small premicellar aggregates are initially formed, and the system reaches equilibrium at around ~ 250 ns (Pires et al., 2012). Experimental observations indicate that the timescale of micellar formation is on the order of 10−6 ~ 10−3 s (Poghosyan et al., 2014). This is too long a timescale for atomistic simulation approaches, but coarse-grained models (Poghosyan et al., 2014) can be used to evaluate the mechanisms of premicellar aggregations. According to these models, SDS premicellar aggregation occurs rapidly by means of a stepwise growth mechanism, i.e. the aggregation number increases gradually via the coalescence of premicellar aggregates. Here, we employ atomistic MD simulations over 1000 ns to probe the denaturation of ACBP in the presence of alkyl sulfates of different chain lengths at the atomic level, and carry out detailed quantitative analyses of the interactions between ACBP and surfactant. We focus on SDS and sodium hexadecyl sulfate (SHS). Thanks to its longer chain length, SHS has a lower cmc than SDS (0.45 versus 8.0 mM in water (Aniansson et al., 1976)) and a higher aggregation number (100 versus 64 (Aniansson et al., 1976)). We analyzed seven independent MD simulations including one control simulation (without any surfactants) and three 1000 ns simulations, each with either 80 SDS or 80 SHS molecules. Our work builds on detailed experimental studies of the unfolding of ACBP in micelles of SDS and sulfates with longer alkyl chains, where we showed that longer alkyl chains lead to faster ACBP unfolding (Andersen and Otzen, 2009). More specifically, the efficiency of denaturation (i.e. the increase in unfolding rate constants with surfactant concentration) scaled in a very consistent manner with the length of the alkyl chain over the range of 8–16 carbon atoms, so that ACBP denaturation was much more sensitive to an increase in the concentration of the surfactants with longer alkyl chains. This scaling behavior suggests, either that the whole alkyl chain interacts with the protein in the denaturation process, or that micelle formation (which scales strictly with chain length) is intimately involved in the denaturation process. A central objective of these simulations is, therefore, to evaluate how micelles and monomeric surfactant molecules interact with ACBP during denaturation. We did not see extensive denaturation of ACBP using preformed micelles but found that a more efficient approach to achieve protein unfolding within our 1-μs window was to incubate ACBP with monomeric surfactant molecules which were able to self-associate and denature ACBP in parallel. To validate our computational approach, MD simulations of ACBP together with SDS are directly compared to experimental data by computing SAXS curves for the ACBP-SDS configurations and comparing them to SAXS data recorded previously (Andersen et al., 2009) and in the current work. Our simulations reveal a common mechanism of denaturation by SDS and SHS, though the process occurs more rapidly with the long-chain surfactant. Partially formed surfactant micelles bind to preferred binding sites on the folded protein that are largely determined by the electrostatic surface potential. These partial micelles grow via the accretion of surfactant monomers from solution and eventually merge to become mature micelles. During this process, the alkyl chains of the surfactant molecules disrupt hydrophobic interactions in the core of the protein. This separates the helices, which gradually lose their hydrogen bonds and become disordered as they are forced to the surface of the micelle. The protein chain is thereby left disordered on the surface of the mature micelle. Simulation Details MD simulations were performed using the GROMACS (version 5.0.4) software package (Berendsen et al., 1995) with GPU support (Abraham et al., 2015). The models of SDS (C12H25SO4Na) and SHS (C16H33SO4Na) were produced in our previous simulations (Poghosyan et al., 2016; Poghosyan et al., 2007). The crystal structure of ACBP was taken from the protein data bank (www.pdb.org, PDB code: 1ACA). ACBP is an all α-helical protein consisting of 86 residues in four helices H1–4 (H1 residues 3–15, H2 20–36, H3 51–62, and H4 65–84) and does not contain any disulfide bonds. The structure was checked for missing atoms, and hetero-atoms were removed. Four independent systems were constructed in which 10, 20, or 80 SDS or 80 SHS molecules were randomly distributed as individual molecules around one ACBP molecule. Based on calculations with the g_mindist module, the minimum distance between surfactant and protein at starting configuration was very similar for both surfactants (0.296 nm for ACBP-SDS and 0.292 nm for ACBP-SHS). The protein and surfactant molecules were inserted into an 8 × 8 × 8 nm3 box containing bulk water using the GROMACS genbox module (Abraham et al., 2015). The number of water (simple point charge (Berendsen et al., 1981)) molecules for both simulations was ~ 15000. This corresponds to an ACBP concentration of 3.24 mM or 30.7 mg/ml. The GROMACS g_density module leads to an ACBP average density of 42 ± 2, 37 ± 2, and 58 ± 2 mg/ml and to an average density of the whole system of 970–990 mg/ml. To ensure that ACBP was stable in the absence of surfactant, we constructed an ACBP/water system and simulated for 1000 ns. All simulations were carried out at constant particle number, pressure, and temperature (NPT ensemble). Enough sodium counterions were added to the simulation box to make the net charge zero. To represent the molecules, we used the GROMACS standard procedure. Thus, for ACBP, we applied pdb2gmx with –ignh, i.e. without explicit aliphatic (non-polar) hydrogens. For SDS and SHS, both methylene and methyl groups were modeled as united atoms. All bonds were kept fixed with the LINCS constraining algorithm (Hess et al., 1987). The temperature was set to 300 K and kept fixed using periodic velocity rescaling (Bussi et al., 2007), where the temperatures of surfactants, protein, counterions, and solvent were controlled independently, employing the v-rescale thermostat algorithm. This generates a proper canonical ensemble which is more ergodic than the Nosé–Hoover approach (Hoover, 1985). The thermostatization on the basis of velocities is preferable, because it is more consistent and avoids the introduction of discontinuities in the generated velocity trajectory (Hünenberger, 2005). Due to the use of cutoff/truncation, the coupling of the whole system to a single thermostat may cause the average solute temperature to be lower than the average solvent temperature. The solution to this problem is to couple separately the solute and solvent degrees of freedom to different thermostats (Hünenberger, 2005). The pressure coupling Berendsen algorithm (Berendsen et al., 1984) was used to keep the system at a constant pressure of 1 atm. The particle mesh Ewald (PME) method was used for long-range electrostatic interactions (Darden et al., 1993), and van der Waals interactions were truncated at 1.4 nm. The equations of motion were integrated using the leapfrog Verlet integrator (Verlet, 1967) with time steps of 2 fs. Three-dimensional periodic boundary conditions were applied, and the coordinates and velocities of all particles were saved for every 0.1 ns. The visualization of the molecular configurations of the system was generated with the VMD package (Humphrey et al., 1996). SS analysis was performed using the Kabsch and Sander method (Kabsch and Sander, 1983) implemented in the do_dssp module provided by GROMACS (Abraham et al., 2015). After system construction, the energy was minimized using steepest descent for 5000 steps to relax high-energy interactions that might have formed during the construction process. From these starting configurations, six MD simulations or replicas, 3 each of 1000 ns (Rep1–3) with either 80 SDS or 80 SHS molecules and an ACBP molecule, were carried out in an NPT ensemble. A control simulation of ACBP over 1000 ns in the absence of surfactants was also performed and analyzed. These simulations were conducted on a computer cluster located at the International Scientific-Educational Center of the National Academy of Sciences of Armenia equipped with GPU Nvidia Tesla graphical processing units (http://bioinformatics.am). The total root mean square deviation of the backbone atoms (RMSD, calculated with the GROMACS g_rms module), root mean square fluctuations of individual residues, number of bound surfactant molecules (within 2.5 Å of any given residue, calculated with the g_mindist module as used elsewhere (Dominguez, 2017)), SS, and hydrogen bonds were analyzed using standard GROMACS tools. To count the number of hydrogen bonds, we use a donoracceptor distance cutoff of 0.35 nm and a cutoff angle of 30° for the angle given by the acceptordonorhydrogen atoms. To characterize the shape of molecular complexes, we calculate the eccentricity derived from the three principal moments of inertia by the following formula: $$\begin{equation} e=1-\frac{I_{\mathrm{min}}}{I_{avg}} \end{equation}$$ (1) where |${I}_{\mathrm{min}}$| is the moment of inertia along the x-, y-, and z-axis with the smallest value and |${I}_{avg}$| is the average of all three moments of inertia. The eccentricity, e, defines the stretching or elongation of the object and is zero for a perfectly spherical object. Preformed micelle construction and simulation details Two preformed micelles (62SDS and 120SHS) were created and were inserted into a water box (8 × 8 × 8 nm3) using the GROMACS genbox module. The corresponding number of sodium counterions was added to neutralize the system. The energy was minimized using steepest descent (5000 steps) to remove unphysical contacts which might have formed during the construction process. Furthermore, after a short NVT equilibration simulation, a production simulation of 100 ns was run. All the bonds were maintained with the LINCS constraining algorithm. The temperature was set to 300 K and kept fixed using periodic velocity rescaling, where the temperatures of surfactants and solvent were controlled independently. The Berendsen pressure coupling algorithm was used to keep the system at a constant pressure of 1 atm. The PME method was used for long-range electrostatic interaction, and van der Waals interactions were truncated at 1.4 nm. The equations of motion were integrated using the leapfrog Verlet integrator with time steps of 2 fs. Three-dimensional periodic boundary conditions were applied, and the coordinates and velocities of all particles were saved for every 0.1 ns. We believe that these are reasonable equilibrium structures of micelles; the approach correctly reproduces dynamical, thermodynamical, and structural properties (Sammalkorpi et al., 2007), including surface tension (Rios-Lopez et al., 2018). Interaction of ACBP with preformed micelles. Construction and simulation details Two systems were constructed, where well-equilibrated micelles with ACBP protein were inserted into water bulk (8 × 8 × 8 nm3). The corresponding number of sodium counterions was added to both systems to neutralize, and furthermore, the energy was minimized using steepest descent (5000 steps) to remove unphysical contacts. After a short NVT run, the systems were subjected to 1000 ns long production runs. All other MD protocols are the same as was described in the previous methods subsections. Experimental Details ACBP was purified as described (Mandrup et al., 1991) and kindly provided as a lyophilized powder by Dr. Kaare Teilum, Copenhagen University. SDS (SigmaUltra) was from Sigma-Aldrich (St. Louis, MO). All experiments were in 10 mM Tris pH 8.0. ACBP was dissolved to a concentration of 1.8 and 3.0 mg/mL (0.2 and 0.32 mM) together with, respectively, 16 and 27 mM SDS. Taking into account a concentration of free SDS of 6 mM as calculated from ITC experiments (Andersen et al., 2009), the SDS/ACBP ratios in the complexes are 56 and 70, respectively. SAXS data was acquired at the flux- and background-optimized Bruker AXS NanoSTAR instrument at Aarhus University (Pedersen, 2004). The instrument uses a liquid metal jet Ga source (Excillum) (Schwamberger et al., 2015), homebuilt scatterless slits, a homebuilt thermostated flow-through capillary, and an automated sample-handler based on Gilson components. Samples were each measured for 900 s, ACBP:SDS 1:56 and ACBP:SDS 1:70, at 20 °C. Buffers were measured as background and water was used for absolute scale calibration (Pedersen, 2004). The intensity is displayed as a function of the modulus of the scattering vector q = (4π/λ)sin(θ), where λ = 1.54 and 1.34 Å is the X-ray wavelength for Cu and Ga sources, respectively, and 2θ is the angle between the incident and scattered X-rays. Snapshot structures were extracted from the simulations for each of the replicas after 200, 400, 600, 800, and 1000 ns. Water was removed and surfactant molecules not connected to the main complex were disregarded. For the calculation of SAXS curves, the excess contrast of protein, surfactant headgroups, and tails were estimated from partial specific volumes. A standard contrast of 2.00 x 1010 cm/g and a representative specific volume of 0.75 cm3/g were used for ACBP. The volumes of the C12 and C16 chains were set to 353 and 462 Å3, respectively, and the headgroup volume was 60.5 Å3 based on published values (Vass et al., 1989). The total excess scattering length density of each of the three components (protein, head, and tail) were evenly distributed on the non-hydrogen atoms of the correponding component. Note that the protein and headgroup have positive excess scattering length, whereas the surfactant tail has a negative excess scattering length. Results and Discussion In what follows, we will describe the mechanism of ACBP denaturation by SDS and SHS, proceeding chronologically from the configurations at the beginning of the simulations, namely the folded state surrounded by randomly oriented surfactant molecules, to the surfactant micelle-bound unfolded state of ACBP present after 1000 ns. Where appropriate, we compare these results to the control simulation of ACBP in water, which contains no surfactants. We also refer to several order parameters that were computed periodically throughout the simulations. These include the root mean squared displacements of the protein backbone atoms (Fig. 1), the SS calculated for each residue (Fig. 2), the fraction of native contacts formed (Q, Fig. 3A and B), the radius of gyration of the protein (Rg,Fig. 3C and D), the distances between the center of mass (COM) of the protein and the COM of each helix (Fig. 4), the number of various types of hydrogen bonds (Fig. 5), the number of surfactant molecules bound within 2.5 Å to the protein (Fig. 6), and the sizes and eccentricities of the surfactant micelles (Fig. 7). We define the native contacts, according to (Best et al., 2013), as a contact between the side chains of two non-neighboring amino acids in the energetically favorable native state of the sequence as defined by the crystal structure of ACBP (PDB: 1ACA). The numbers of protein-protein, protein-water, and surfactant-protein hydrogen bonds were separately enumerated. The size and shape are quantified using the radii of individual surfactant aggregates and the eccentricities of the ACBP/micelle complexes. Fig. 1 Open in new tabDownload slide Backbone RMSD of particles of ACBP either alone (blue) or in the presence of surfactant (3 replica runs Rep1–3) as a function of time: (A) ACBP-SDS and (B) ACBP-SHS. Note that for the control simulation, the relaxation phase over the first 100 ns reflects adjustment from the previous NVT run. Fig. 1 Open in new tabDownload slide Backbone RMSD of particles of ACBP either alone (blue) or in the presence of surfactant (3 replica runs Rep1–3) as a function of time: (A) ACBP-SDS and (B) ACBP-SHS. Note that for the control simulation, the relaxation phase over the first 100 ns reflects adjustment from the previous NVT run. Fig. 2 Open in new tabDownload slide Time evolution of SS. (A) ACBP in water. (B) ACBP-SDS Rep1–3. (C) ACBP-SHS Rep1–3. SS was assigned by the GROMACS do_dssp module (Abraham et al., 2015). Fig. 2 Open in new tabDownload slide Time evolution of SS. (A) ACBP in water. (B) ACBP-SDS Rep1–3. (C) ACBP-SHS Rep1–3. SS was assigned by the GROMACS do_dssp module (Abraham et al., 2015). Fig. 3 Open in new tabDownload slide The fraction of native contacts (A, B) and the radius of gyration (C, D) of ACBP in the presence of SDS (A, C) and SHS (B, D) as a function of time. The control represents ACBP gyration data from control simulation (ACBP in water). The stippled line is the average of the 3 replicate runs. Fig. 3 Open in new tabDownload slide The fraction of native contacts (A, B) and the radius of gyration (C, D) of ACBP in the presence of SDS (A, C) and SHS (B, D) as a function of time. The control represents ACBP gyration data from control simulation (ACBP in water). The stippled line is the average of the 3 replicate runs. Fig. 4 Open in new tabDownload slide The distances of COM of each helix from the protein COM (A) in water and in the presence of (B) SDS and (C) SHS. The three different replicas are in the same color or shade gray as in panel A. Note the same y-axis in all three panels. Fig. 4 Open in new tabDownload slide The distances of COM of each helix from the protein COM (A) in water and in the presence of (B) SDS and (C) SHS. The three different replicas are in the same color or shade gray as in panel A. Note the same y-axis in all three panels. Fig. 5 Open in new tabDownload slide (A) and (B) The number of main chain–main chain hydrogen bonds within ACBP (Rep1–3, blue) and between ACBP and surfactant (Rep1–3, red). Panel A has ACBP-SDS and Panel B ACBP-SHS data. (C) The number of protein-water hydrogen bonds in ACBP-SDS (green) and ACBP-SHS (orange). (D) The number of main chain–main chain hydrogen bonds within ACBP in water. Fig. 5 Open in new tabDownload slide (A) and (B) The number of main chain–main chain hydrogen bonds within ACBP (Rep1–3, blue) and between ACBP and surfactant (Rep1–3, red). Panel A has ACBP-SDS and Panel B ACBP-SHS data. (C) The number of protein-water hydrogen bonds in ACBP-SDS (green) and ACBP-SHS (orange). (D) The number of main chain–main chain hydrogen bonds within ACBP in water. Fig. 6 Open in new tabDownload slide The number of bound (A) SDS and (B) SHS molecules (within 2.5 Å of any residue on ACBP) as a function of simulation time. Fig. 6 Open in new tabDownload slide The number of bound (A) SDS and (B) SHS molecules (within 2.5 Å of any residue on ACBP) as a function of simulation time. Fig. 7 Open in new tabDownload slide The radii of surfactant aggregates as a function of time for both (A) ACBP/SDS (Rep1) and (B) ACBP/SHS (Rep1) simulations. Insets: A plot of the eccentricity of the surfactant-ACBP complex as a function of time. Different colors or shades of gray refer to different aggregates which merge during the simulation. Fig. 7 Open in new tabDownload slide The radii of surfactant aggregates as a function of time for both (A) ACBP/SDS (Rep1) and (B) ACBP/SHS (Rep1) simulations. Insets: A plot of the eccentricity of the surfactant-ACBP complex as a function of time. Different colors or shades of gray refer to different aggregates which merge during the simulation. First 50 ns: surfactant cluster binding and partial unfolding When ACBP is incubated with SDS, small clusters of 10–20 SDS molecules with the radii of 1.0–1.6 nm (Fig. 7) form in solution within the first 3–5 ns (Fig. 8). In parallel, smaller clusters of 3–5 surfactant monomers begin to interact with the folded ACBP molecule (Fig. 6), binding in numerous small steps. Note that the apparent sharp jumps in the number of bound molecules is a technical artifact resulting from the binning of individual frames to reduce the size of the output files. Taking a detailed look at the first replica of the ACBP-SDS system as an example, the first close contact between SDS and ACBP is observed at ~ 0.8 ns, where four SDS molecules bind to the Val12-Asp21 segment of ACBP, linking helices I and II (Fig. 8A). This segment contains two Lys residues (Lys13 and Lys18), both of which are part of the ligand binding site of ACBP (Kragelund et al., 1993). These four bound SDS molecules reorient over the next 0.2–0.3 ns such that the SDS head groups maintain contact with Val12-Lys16 and the alkyl tails approach the hydrophobic region around the termini of helices I and II (Fig. 8B–D). Over the next 2 ns, two of these SDS molecules migrate to H4 near the Ala72-Gly85 region (Fig. 8E) and 8–9 additional SDS molecules bind, covering almost one entire side of ACBP (Fig. 8F). At this stage, however, ACBP remains folded in the native state (RMSD < 0.4 nm, Fig. 1, Q > 0.7, Fig. 3A). Interestingly, the Lys residues in the active site seem to bind SDS molecules preferentially. After 5 ns, four of the five active site Lys residues have SDS molecules in close contact (<2.5 Å), whereas this is only the case for one of the 10 non-active site Lys (Fig. 9A and C). The SDS molecules contact Lys mainly through sulfate-amine contacts while clustering their alkyl chains through intermolecular SDS–SDS contacts (Fig. 9B). In general, for the ACBP-SDS simulations, only a few (8–20) SDS monomers bind ACBP within the first 5–10 ns, and this does not immediately lead to any significant disruption of SS, loss of internal hydrogen bonds, or decrease in the fraction of native contacts, Q. Fig. 8 Open in new tabDownload slide Snapshots extracted from the beginning of the first simulation run for ACBP/SDS (Rep1). (A) 0.8 ns. (B) 0.85 ns. (C) 0.9 ns. (D) 1.0 ns. (E) 1.4 ns. (F) 3.3 ns. SDS molecules are represented with the following atom colors: sulfur—yellow, carbon—cyan (free SDS), or red (bound SDS). Water and sodium counterions are omitted for clarity. The snapshots were rendered using VMD. Fig. 8 Open in new tabDownload slide Snapshots extracted from the beginning of the first simulation run for ACBP/SDS (Rep1). (A) 0.8 ns. (B) 0.85 ns. (C) 0.9 ns. (D) 1.0 ns. (E) 1.4 ns. (F) 3.3 ns. SDS molecules are represented with the following atom colors: sulfur—yellow, carbon—cyan (free SDS), or red (bound SDS). Water and sodium counterions are omitted for clarity. The snapshots were rendered using VMD. Fig. 9 Open in new tabDownload slide Binding of SDS and SHS around the active site. The fraction of different residue classes within 2.5 Å of (A) SDS and (D) SHS molecules at 12 different time points during the run. (B) Snapshot of binding of SDS after 5 ns, where four of the five active site Lys residues are in contact with SDS. (C) Snapshots of binding of SHS to ACBP after 0.5, 0.7, and 1.1 ns, respectively, illustrating early clusters formed around residues 44–56 and H4. Fig. 9 Open in new tabDownload slide Binding of SDS and SHS around the active site. The fraction of different residue classes within 2.5 Å of (A) SDS and (D) SHS molecules at 12 different time points during the run. (B) Snapshot of binding of SDS after 5 ns, where four of the five active site Lys residues are in contact with SDS. (C) Snapshots of binding of SHS to ACBP after 0.5, 0.7, and 1.1 ns, respectively, illustrating early clusters formed around residues 44–56 and H4. For SHS, the results are broadly similar, with two major differences: binding is faster and more cooperative. The clusters of surfactant molecules interacting with folded ACBP are larger and encompass ~ 15 molecules (Fig. 6). Consistent with this, SHS molecules bind in fewer steps than SDS with up to ~ 20 monomers binding almost simultaneously. After ca. 0.5 ns, SHS clusters of 4–6 molecules have formed near residues ~ 44–56 and the H4 helix, and these aggregates grow in size over the next few nanoseconds (Fig. 9C). Early binding also targets the ligand binding site, leading to contact with three out of five active-site Lys after 3 ns while no other Lys are targeted in this period (Fig. 9D). These contacts involve the clusters of SHS that are even larger than those of SDS. In contrast to SDS, binding of these and other clusters leads to an increase of the RMSD values to around 0.8 nm and a drop in the fraction of native contacts, Q, to 0.55–0.6 in all replicas. It is instructive to compare these values to the corresponding values in the control simulation, which fluctuate solely as the result of thermal fluctuations and interactions with the solvent (Fig. 3). For example, the moderate rise in RMSD seen in the SDS simulations is matched by the control simulation (Fig. 1A), whereas all of the SHS simulations clearly deviate more from the starting structure than the control (Fig. 1B). The initial fluctuations in the fraction of native contacts, Q, are also similar between the control simulation and the SDS simulations, but Q quickly rebounds in the case of the control, which is indicative of fluctuations around an average (and well-folded) structure. Indeed, the RMSD values in the control simulation were confined to less than 0.4 nm for the duration of the 1000 ns simulation. The simulations with SDS and SHS, on the other hand, show nearly monotonic increases in RMSD and steadily decreasing Q values (Fig. 1). The most robust order parameter for distinguishing between conformations populated in the control and experimental conditions during the first 50 ns is the protein’s radius of gyration Rg (Fig. 3). All simulations with surfactants show a distinct increase in Rg within a few nanoseconds, whereas the value in the control simulation fluctuates between 1.2 and 1.3 nm. The early stages of denaturation by surfactant evidently involve a significant expansion of the native structure. This expansion is more pronounced and faster in the case of SHS as compared with SDS. For SHS, an expansion from 1.3 to 1.7 nm in Rg (corresponding to a ~ 120% volume increase) is seen in all replica simulations within 50 ns. In the period 50–250 ns, we see growth of ACBP-bound surfactant micelles and binding of additional partial micelles, tertiary structure distortion via penetrating alkyl chains, and the replacement of protein-protein hydrogen bonds with surfactant-protein bonds. Following initial surfactant cluster binding and partial unfolding, over the next ~ 200 ns, the ACBP-surfactant complex undergoes a slower phase of simultaneous ACBP-bound micelle growth and further loss of tertiary contacts. Binding of secondary clusters of SDS and SHS molecules occurs largely within 50–100 ns (Fig. 6), after which the number of bound surfactant molecules remains approximately constant. In contrast, the shape of the ACBP-surfactant complex undergoes a continuous rearrangement up to 300–400 ns, leading to a steady decrease of the eccentricity of the ACBP-surfactant complex from 0.6 to 0.2–0.3 for SDS and 0.1–0.2 for SHS (Fig. 7, insets). Lower eccentricities indicate a more spherical shape, making the ACBP-SHS complex more spherical than its ACBP-SDS counterpart. Intuitively, one might expect a larger alkyl chain to alter protein shape to a greater extent; however, this is not the case. In fact, changes in chain length do not lead to changes in micelle shape distributions (Iyer and Blankschtein, 2012; Oliver et al., 2013), in contrast to, e.g. changes in ionic strength (Jönsson et al., 1998). Reorganization of the surfactant micelles in bulk solution is also ongoing during this phase as they merge into fewer and larger micelles in multiple steps. During this phase, we see the first significant disruption of secondary structural elements, particularly with SHS (Fig. 2). As a result of the tertiary rearrangements, Q decreases steadily and Rg rises, indicating a further expansion of the structure as (mostly tertiary) native contacts are broken. In particular, protein-protein hydrogen bonds are lost, decreasing from ~ 50 to 30 over 250 ns, whereas the number of protein-protein hydrogen bonds in the control simulation fluctuates between 42 and 55 (Fig. 5). In the presence of surfactant, the broken protein-protein hydrogen bonds are replaced with an approximately equal number of surfactant-protein hydrogen bonds, i.e. the sulfate group of SDS efficiently disrupts the intramolecular hydrogen bonds and replaces them with hydrogen bonds between SDS sulfate oxygen atoms and protein amide nitrogen atoms as seen in other simulations (Poghosyan et al., 2014). The number of protein-water hydrogen bonds in the presence of surfactant shows no distinct trend as compared with the control simulation, and there is a conservation of the hydrogen bond length distribution irrespective of the type of surfactant (data not shown). Although the trends in the number and type of hydrogen bonds over time are qualitatively similar in the SDS and SHS simulations, the replacement of protein-protein hydrogen bonds with protein-surfactant hydrogen bonds is both more rapid and more complete in SHS (Fig. 5B) than in SDS (Fig. 5A), and SHS also leads to a larger number of ACBP-water bonds than SDS (Fig. 5C). Apparently, SHS’s longer alkyl chain enhances its ability over SDS to disrupt internal hydrogen bonds despite the two surfactants having identical head groups. Overall, the different replica runs follow the same trajectory. For illustrative purposes, we take a closer look at Rep1, the first replica of the ACBP-SHS system between 50 and 500 ns (Fig. 10A–C). Small clusters of SHS molecules aggregate on ACBP over the first ~100 ns nanoseconds (Fig. 10A and B), but the first micelle attack happens at ~ 170 ns, when a micelle aggregate with 27 SHS monomers binds to H3 around residues Lys52-Leu61 (blue micelle in Fig. 10C, see timeline in Fig. 6B), probably aided by the charge complementarity of the Lys residues at positions 50, 52, and 54. When the micelle attacks ACBP, the remaining helical content (H3 and H2) disappears rapidly for H3 and more slowly for H2, which still has some helical structure at 350 ns (Fig. 10D). Finally, all helical content disappears around ~ 500 ns (Figs 2 and 10E). Binding of additional micelles from solution to the already partially denatured ACBP structure quickly leads to further denaturation. The other ACBP-SHS runs gave broadly similar trajectories although there were minor variations in the exact times of the different events and the number of bound SHS monomers. Fig. 10 Open in new tabDownload slide Snapshots extracted over 500 ns for ACBP/SHS Rep1. (A) 20 ns. (B) 130 ns. (C) 170 ns. (D) 350 ns. (E) 500 ns. Atoms of protein-bound SHS molecules are represented as follows: sulfur—yellow and carbon—cyan. Blue and green atoms represent sulfur atoms of free micelles. Water and sodium counterions are omitted for clarity. The snapshots were rendered using VMD package. Fig. 10 Open in new tabDownload slide Snapshots extracted over 500 ns for ACBP/SHS Rep1. (A) 20 ns. (B) 130 ns. (C) 170 ns. (D) 350 ns. (E) 500 ns. Atoms of protein-bound SHS molecules are represented as follows: sulfur—yellow and carbon—cyan. Blue and green atoms represent sulfur atoms of free micelles. Water and sodium counterions are omitted for clarity. The snapshots were rendered using VMD package. A look at the distances between the entire protein’s center of mass and the individual helices’ centers of mass reveals important differences between the two surfactants. In the case of SDS, these distances, which start at different values because of the different positions of the helices in the native structure, show a nearly uniform increase (Fig. 4B), i.e. the helices uniformly expand away from the COM of the protein. Thus, 250 ns of simulations with SDS lead to a distorted and expanded version of the native state. SHS, in contrast, leads to a collapse toward a single value (Fig. 4C), i.e. the helices have lost almost all memory of their relative placement in the folded state and the structure is thoroughly denatured. Both cases differ dramatically from the control simulation (Fig. 4A), wherein the native topology is retained and, therefore, only minor deviations from the starting distances occur throughout the simulation. In the period 250–500 ns, unraveling of relatively hydrophilic helices occurs at the micelle–solvent interface, whereas hydrophobic helices persist through interactions with surfactant alkyl chains. Most major disruptions to the native SS occur between 100 and 500 ns. In general, there is a high degree of cooperativity in the helical structures (Fig. 2). Within the first 300 ns of simulation, H1 is converted from being entirely helical to a collection of bend and turn structures, and this happens faster, on average, in the presence of SHS as compared with SDS. After H1, H3 is the next-most prone to disruption, showing complete or partial disruption in the presence of SDS, and is invariably completely disrupted in the presence of SHS within 400 ns. H4 is largely retained in SDS but shows complete denaturation in two out of the three SHS simulations. H2 is the most resistant to disruption. The least hydrophobic helices, H1 and H3, tend to unravel first, whereas the most hydrophobic helix, H2, is preserved in all six surfactant simulations with only one exception: Rep1 of the SHS simulations. In this case, H2 unravels but, suggestively, this uncommon event is coincident with another unique event in our current set of simulations, namely the conversion of part of H2 into a β-strand through interaction with an N-terminal segment of H4. More hydrophobic helices tend to interact preferentially with the alkyl chains of surfactant molecules, which form a poor environment for the exchange of hydrogen bonds. This trend explains the tendency for the SDS-denatured state of membrane proteins, whose transmembrane helices tend to be very hydrophobic, to contain intact helices (Krishnamani et al., 2012). Helices in soluble proteins that are amphipathic or hydrophilic will interact preferentially with solvent and the sulfate head groups of surfactant molecules and, therefore, readily undergo interchange of hydrogen bonds, which ultimately results in the disruption of helical SSs. These disruptions in SS are reflected in steadily decreasing Q values and increasing Rg (Fig. 3) The radius of gyration of SDS-denatured ACBP rises approximately linearly up to ~ 1.7–1.85 nm at around 450–500 ns and thereafter remains relatively constant. ACBP reaches this Rg value more quickly in the presence of SHS and changes less than in the presence of SDS between 100 and 500 ns. The overall view of denaturation is that multiple ACBP-bound micelles coalesce and ACBP denatures completely as it migrates to the surface of a mature surfactant micelle and loses almost all remaining native structure. In general, initial binding of surfactant molecules to ACBP occurs via clusters whose size is determined by the average cluster size present in solution, which is smaller in the case of SDS at the start of the simulation. These ACBP-bound clusters do not dissociate from ACBP in our simulations; instead, they grow via accretion of surfactant monomers and clusters from solution. The resulting structure destabilization and diffusion of free surfactant clusters leads to new binding events between the partially unfolded structure of ACBP and partial surfactant micelles from bulk solution. As these multiple bound clusters grow, their propensity to coalesce while bound to the protein increases, and the eventual coalescence of these partial micelles results in the formation of mature surfactant micelles. These coalescence events coincide with large disruptions to ACBP’s structure. Parts of the protein chain that were on the interface between these partial micelles prior to coalescence are pushed to the surface of the newly merged micelle. Unlike the other helices in ACBP, which get pushed to the surface and unravel when protein-bound micelles coalesce, the most hydrophobic helix, H2, persists by preferentially interacting with the alkyl chains of the surfactant molecules. In the hydrophobic environment, of the micelle interior, the helical hydrogen bonds have no ‘good alternatives’, and we only see the full disruption of H2 when the helical hydrogen bonds are replaced by β-strand hydrogen bonds via interactions with a different (relatively hydrophobic) part of the protein chain. The end result of the overall mechanism described above is a largely disordered ACBP chain (lacking all tertiary structure and most helices) that wraps around a mature surfactant micelle. This is sometimes referred to as a ‘flexible capped micelle’ (Lundahl et al., 1986), where it is thought that binding via complementary charges is an important contributor to the stability of the complex. The reduction in native-level SS is supported by the previous far-UV circular dichroism spectra of ACBP in the absence and presence of 10 mM SDS (Andersen et al., 2009), which shows a marked shift in the relative magnitudes of the minima at 208 and 220 nm. Deconvolution of these spectra by BestSel (Micsonai et al., 2015) shows a decrease in helicity from ca. 54% to around 35% in the presence of 10 mM SDS (data not shown). Comparison with other surfactant systems for denaturation of ACBP To denature ACBP in surfactants in silico, our approach has been to incubate one ACBP molecule with 80 individual surfactant molecules rather than micelles. ACBP is denatured by SDS and SHS in vitro at concentrations well below the cmc (6 mM for SDS and 0.38 mM for SHS (Andersen and Otzen, 2009)), which means that no micelles are present in the solution from the onset of the experiment though clusters can subsequently form on the protein. In silico, it is not feasible to operate at such low surfactant concentrations, since the current system volume of 8 × 8 × 8 nm3 leads to a protein concentration (at one protein molecule per cell) of 6.5 mM, so that surfactant-protein ratios of 80:1 require 520 mM surfactants. The very high surfactant concentrations combined with the far-from-equilibrium initial configuration of randomly placed surfactant molecules around a well-folded protein structure leads to micelle formation occurring in parallel with protein denaturation. This somewhat unphysical starting point for the simulations raises the concern that the observed denaturation pathway largely an artifact of the relaxation of a system that is initialized far from equilibrium. To begin to address this concern, we also carried out two alternative simulations, namely (i) reducing the number of surfactant molecules from 80 to 20 and 10 (to test whether lower concentrations would still lead to ACBP unfolding) and (ii) incubation with preformed micelles of SDS and SHS. Both using preformed micelles and using a smaller number of monomers resulted in ACBP being less unfolded at the end of a 1000 us simulation than was seen in our original 80-monomer simulations (Fig. 11). In the case of SDS, the preformed micelle (Fig. 11A, top) only led to a very modest decline in the number of native contacts, whereas the inclusion of 10 and 20 molecules led to a slightly larger (but still incomplete compared to the 80-molecule simulation) reduction in native structure. Incubation with a preformed SHS micelle (Fig. 11B, top) led to an even more modest decline in native contacts which reached a plateau within a few hundred nanoseconds, reflecting the unfolding of helix 1. In both cases, ACBP remains bound to the surface of the micelle with most helices intact apart from helix 1, although helix 3 seems to become more dynamic in SHS. Due to the kinetic barriers associated with repartitioning of aliphatic chains from the micelle interior to the protein, initializing the simulations using randomly placed monomeric surfactant leads to comparatively rapid equilibration of the surfactant-protein system. In the following sections, we compare experimental SAXS data to scattering curves computed using the structures obtained at the end of the simulations that were initialized with monomeric surfactant, where the protein structure is largely denatured as is expected at equilibrium. Fig. 11 Open in new tabDownload slide Simulations of ACBP denaturation initialized using either pre-formed micelles, 10 SDS monomers, or 20 SDS monomers. The two panels show the fraction of native contacts over time as ACBP is incubated with either preformed micelles of (A) SDS or (B) SHS or with different numbers of surfactant monomers as indicated in the legends. Snapshots of the equilibrated micelles of SDS (62 monomers) and SHS (120 monomers) are shown above the graphs. (C) Snapshots from various simulations after 1000 ns. (D) Time evolution of secondary structure for the indicated complexes between ACBP and micelles or smaller numbers of surfactant molecules. Fig. 11 Open in new tabDownload slide Simulations of ACBP denaturation initialized using either pre-formed micelles, 10 SDS monomers, or 20 SDS monomers. The two panels show the fraction of native contacts over time as ACBP is incubated with either preformed micelles of (A) SDS or (B) SHS or with different numbers of surfactant monomers as indicated in the legends. Snapshots of the equilibrated micelles of SDS (62 monomers) and SHS (120 monomers) are shown above the graphs. (C) Snapshots from various simulations after 1000 ns. (D) Time evolution of secondary structure for the indicated complexes between ACBP and micelles or smaller numbers of surfactant molecules. Experimental SAXS data validate simulation results To link our simulations with experimental measurements, we turned to SAXS. Other than its neutron counterpart SANS, SAXS is the only experimental method that can provide structural information on shape, size, and the aggregation number of micelles, and protein-surfactant complexes. SAXS is particularly sensitive to the structure of these complexes as the micelles show a positive excess scattering length density for the headgroup shell and a negative excess scattering length for the core, allowing core and shell to be distinguished. For the core-shell complexes, we use the absolute scattering intensity which provides the protein mass per complex. Also, for both micelles and complexes, SAXS signal intensity is very sensitive to overall symmetry of the particles. The minimum in the intensity observed between the main intensity at low values of the scattering vector q and the secondary maximum is only deep and well-defined when the structures have a well-defined structure with a homogeneously distribution of excess scattering length density in the shell. We have confirmed this in calculations for core-shell ellipsoids, where the center of core and shell are displaced relative to each other (J.S.P., unpublished observations). For our comparison, we used experimental spectra with the same ACBP-surfactant composition as the simulations, using spectra either available previously (in the case of a complex with an SDS:ACBP molar ratio of 40:1 (Andersen et al., 2009)) or recorded for the present state (SDS:ACBP molar ratios of 56 and 70). The experimental spectra in Fig. 12A and B for complexes with about 56 and 70 SDS molecules display deep minima in the intensity at approximately q = 0.13 Å−1 in both data sets, and a broad subsidiary maximum at higher q. Such a q-dependence is the characteristic of scattering from core-shell structures with a high degree of centro-symmetry for which the excess scattering length density has opposite signs in core and shell. Our previous SAXS data for ACBP complexes with about 40 SDS molecules per ACBP molecule (Andersen et al., 2009) have similar deep minima and thus also high symmetry. Fig. 12 Open in new tabDownload slide Experimental SAXS data for the samples with (A) 56 and (B) 70 SDS molecules in the ACBP-SDS complexes. The curves are scaled to the experimental data, and a structure factor taking into account inter-complex repulsion is also optimized. The line shows the SAXS curves calculated for the simulations at the end of the runs (1000 ns) for Rep1 and Rep3 which contain 64 and 80 SDS molecules, respectively, bound to ACBP. Fig. 12 Open in new tabDownload slide Experimental SAXS data for the samples with (A) 56 and (B) 70 SDS molecules in the ACBP-SDS complexes. The curves are scaled to the experimental data, and a structure factor taking into account inter-complex repulsion is also optimized. The line shows the SAXS curves calculated for the simulations at the end of the runs (1000 ns) for Rep1 and Rep3 which contain 64 and 80 SDS molecules, respectively, bound to ACBP. Fig. 13 Open in new tabDownload slide SAXS curves calculated for the complexes during the simulations. Panels A–C: SDS Rep1–3. Panels D–E: SHS Rep1–3. Curves are for 200 ms (grey, full), 400 ms (grey, dashed), 600 ms (black, dotted), 800 ms (black, dashed), 1000 ms (black, full). Fig. 13 Open in new tabDownload slide SAXS curves calculated for the complexes during the simulations. Panels A–C: SDS Rep1–3. Panels D–E: SHS Rep1–3. Curves are for 200 ms (grey, full), 400 ms (grey, dashed), 600 ms (black, dotted), 800 ms (black, dashed), 1000 ms (black, full). The experimental SAXS data were then compared with the SAXS curves calculated for the structures present at the end of the simulations (Fig. 12). An overall scale factor was used to optimize the agreement. The data have the signature of concentration effects (interparticle repulsion) at low q with a decrease of intensity. This was modeled by including an effective hard-sphere structure factor, which includes a hard-sphere volume fraction and hard-sphere interaction radius that were also optimized in the fit. The best fits were found with Rep1 for the data set with 56 SDS molecules and for Rep3 for the sample with 70 SDS molecules (Fig. 12). The corresponding number of SDS molecules for the simulations is 64 and 80, respectively, at the end of the 1000-ns runs (Fig. 6A). As the SAXS curves are particularly sensitive to the symmetry of the formed structures, the calculated curves during complex formation can provide an impression of this and whether complex formation and growth is a monotonic process. Figure 13 shows the calculated SAXS curves for both SDS and SHS at different time points. All results are in agreement with a core-shell structure as the SAXS curves have clear oscillations. For both SDS and SHS, the results display a quite large variation due to the stochastic nature of the simulations. Furthermore, there is also not a monotonic growth and behavior, except for the third run for SDS (Fig. 13C). For the second run for both SDS and SHS, the structure after 200 ns has particularly low symmetry. For SDS, the first minimum is not very deep for any of the calculated curves, in contrast to the situation for the two first runs for SHS. This could suggest that the driving force for forming highly symmetric structures is larger for SHS due to the larger hydrophobicity of long hydrocarbon chains of this surfactant. Comparison with experimental data highlights the dynamics of helix H1 and the key role of Lys residues in the active site region Examining changes in SS in the simulations provides a particularly fruitful point of comparison with existing experimental data. The first unfolding event in our simulations takes place after ~ 100 ns with SDS and after ~ 70 ns with SHS. H1 is the first site of unfolding for both surfactants, but it needs 100 ns to unfold in SDS, while 20 ns in SHS is sufficient to unravel half of the helix and an additional 60 ns leads to complete unfolding. This is in excellent agreement with experimental measurements which show that H1 is significantly more dynamic than other helices (Kragelund et al., 1995). In general, more hydrophobic helices appear to be more resistant to disruption. We attribute this to their preference for interacting with the alkyl chains of surfactant molecules; this makes it less favorable for them to access the myriad alternative hydrogen bonding arrangements that are available to helices that partition to the surface of the micelle and interact with solvent molecules and the polar surfactant head groups. This observation offers a prediction that could be experimentally validated. Indeed, this is already in agreement with experimental data on SDS-denatured membrane proteins that show native-like levels of helical SS (Paslawski et al., 2015). The overall increased aggressiveness of SHS is in good agreement with experimental results which reveal that even when corrected for its lower cmc, SHS more efficiently denatures ACBP than SDS (Andersen and Otzen, 2009), indicating that SHS binds even better to ACBP than to itself. Another remarkable observation is the role played by positively charged residues around the active site. ACBP has no Arg but 15 Lys residues; of these, five (13, 18, 32, 50, and 54) are quite close to the active site. Based on the aberrant behavior of mutations in this region (Andersen and Otzen, 2009), the active site was previously identified as a possible site for early attack. Closer inspection provides a remarkable vindication of this prediction. When the proximity of SDS to Lys residues is plotted versus time, it is abundantly clear that the five active site Lys residues are saturated with SDS contacts (defined as being within < 2.5 Å of an SDS molecule) much earlier than the other 10 Lys residues and the remaining residues (Fig. 9). This precocious affinity for SDS seems to be connected with the combination of favorable electrostatic interactions to the numerous Lys residues in the ligand binding site, combined with the ability to sequester the alkyl chains through intermolecular surfactant–surfactant interactions. Conclusions Using atomistic MD simulations, we have studied the surfactant-induced denaturation of the protein ACBP. Our simulations are distinguished by the high level of atomic resolution and the duration of the trajectories, allowing us to compare and contrast the effect of two different surfactants on the denaturation of ACBP and to gain insight into the microscopic basis of the denaturation mechanism. We have previously established that increasing surfactant chain length increases the rate of denaturation (Andersen and Otzen, 2009). This is in good agreement with our simulations. SHS unfolds ACBP more quickly and efficiently than SDS, leading to a nearly complete loss of helical structure. This process occurs in several steps: initial binding of small surfactant clusters leads to an expansion of the native state but retention of most native structure, particularly in the case of SDS. Growth of these clusters via accretion of surfactant monomers and clusters from solution leads to the disruption of hydrophobic interactions by the surfactant alkyl tails and the replacement of protein-protein hydrogen bonds by protein-surfactant hydrogen bonds with the surfactant head groups. Partial structure destabilization leads not only to the growth of ACBP-bound surfactant micelles but also to the further binding of micelles from solution. As these bound micelles grow and eventually merge, the disordered protein chain migrates to the surface of the micelle. The resulting ACBP-surfactant complexes are, in general, ellipsoidal, although the ACBP-SHS complexes are nearly spherical when the denaturation of ACBP is complete. The high symmetry of the ACBP-SHS complexes are confirmed by experimental SAXS curves that showed pronounced oscillations as expected for core-shell structures with a symmetric distribution of the protein around the micelle. This work provides further atomistic insight into the mechanism of surfactant-induced protein denaturation, and we hope that it will inspire additional experimental endeavors to test and add even more depth to our conclusions. Acknowledgements We acknowledge supercomputer time provided at the Institute for Informatics and Automation Problems of NAS RA, Armenia. We are grateful to Dr. Kaare Teilum for the generous gift of purified ACBP. Funding This work was supported by the EC [HORIZON2020–675121–VI-SEEM to A.H.P.]; The Danish Research Council|Natural Sciences [4090–00220 to N.S. and D.E.O]; the Carlsberg Foundation [CF14–0287 to N.S. and D.E.O]. References Abraham , M.J. , Murtola , T. , Schulz , R. , Pall , S. , Smith , J.C. , Hess , B. , Lindahl , E. ( 2015 ) SoftwareX , 1 , 19 – 25 . Crossref Search ADS Aehle , W. (ed) ( 2007 ) Enzymes in Industry. Production and applications . Wiley , Weinheim . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Andersen , K.K. , Oliveira , C.L.P. , Larsen , K.L. , Poulsen , F.M. , Callisen , T.H. , Westh , P. , Pedersen , J.S. , Otzen , D.E. ( 2009 ) J. Mol. Biol. , 391 , 207 – 226 . Crossref Search ADS PubMed Andersen , K.K. and Otzen , D.E. ( 2009 ) J. Phys. Chem. B , 113 , 13942 – 13952 . Crossref Search ADS PubMed Aniansson , E.A.G. , Wall , S.N. , Almgren , M. , Hoffmann , H. , Kielmann , I. , Ulbricht , W. , Zana , R. , Lang , J. , Tondre , C. ( 1976 ) J. Phys. Chem. , 80 , 905 – 922 . Crossref Search ADS Bales , B.L. and Algren , M. ( 1995 ) J. Phys. Chem. , 99 , 15153 – 15162 . Crossref Search ADS Berendsen , H.J.C. , Postma , J.P.M. , Van Gunsteren , W.F. , DiNola , A. , Haak , J.R. ( 1984 ) J. Chem. Phys. , 81 , 3684 – 3690 . Crossref Search ADS Berendsen , H.J.C. , Postma , J.P.M. , van Gunsteren , W.F. , Hermans , J. ( 1981 ) Intermolecular Forces , Pullman , B. (ed). Reidel , Dordrecht , pp. 331 – 342 . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Berendsen , H.J.C. , van der Spoel , D. , van Drunen , R. ( 1995 ) Comput. Phys. Commun. , 91 , 43 – 56 . Crossref Search ADS Bergström , M. and Pedersen , J.S. ( 1999 ) Phys. Chem. Chem. Phys. , 1 , 4437 – 4446 . Crossref Search ADS Berr , S.S. and Jones , R.R.M. ( 1988 ) Langmuir , 4 , 1247 – 1251 . Crossref Search ADS Best , R.B. , Hummer , G. , Eaton , W.A. ( 2013 ) Proc. Natl. Acad. Sci. U. S. A. , 110 , 17874 – 17879 . Crossref Search ADS PubMed Bussi , G. , Donadio , D. , Parrinello , M. ( 2007 ) J. Chem. Phys. , 126 , 014101 . Crossref Search ADS PubMed Chen , J.-M. , Su , T.-M. , Mou , C.Y. ( 1986 ) J. Phys. Chem. , 90 , 2418 – 2421 . Crossref Search ADS Darden , T. , York , D. , Pedersen , L. ( 1993 ) J. Chem. Phys. , 98 , 10089 . Crossref Search ADS Dominguez , H. ( 2017 ) J. Mol. Model. , 23 , 210 . Crossref Search ADS PubMed Hess , B. , Bekker , H. , Berendsen , H.J.C. , Fraaije , J. ( 1987 ) J. Comput. Chem. , 18 , 1463 – 1472 . Crossref Search ADS Hoover , W.G. ( 1985 ) Phys Rev A Gen Phys , 31 , 1695 – 1697 . Crossref Search ADS PubMed Humphrey , W. , Dalke , A. , Schulten , K. ( 1996 ) J. Mol. Graph. , 14 , 33 – 38 . Crossref Search ADS PubMed Hünenberger , P.H. ( 2005 ) Adv. Polym. Sci. , 173 , 105 – 149 . Crossref Search ADS Højgaard , C. , Sørensen , H.V. , Pedersen , J.S. , Winther , J.R. , Otzen , D.E. ( 2018 ) Biophys. J. , 115 , 2081 – 2086 . Crossref Search ADS PubMed Iyer , J. and Blankschtein , D. ( 2012 ) J. Phys. Chem. B , 116 , 6443 – 6454 . Crossref Search ADS PubMed Jensen , G.V. , Lund , R. , Gummel , J. , Narayanan , T. , Pedersen , J.S. ( 2014 ) Angew. Chem. Int. Ed. Engl. , 53 , 11524 – 11528 . Crossref Search ADS PubMed Jones , M.N. ( 1996 ) In Magdassi , S. (ed)., Surface Activity of Proteins . New York : Marcel Dekker , pp. 237 – 284 . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Jönsson , B. , Lindman , B. , Holmberg , K. , Kronberg , B. ( 1998 ) Surfactants and Polymers in Aqueous Solutions . Wiley & Sons , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Kabsch , W. and Sander , C. ( 1983 ) Biopolymers , 22 , 2577 – 2637 . Crossref Search ADS PubMed Kragelund , B.B. , Andersen , K.V. , Madsen , J.C. , Knudsen , J. , Poulsen , F.M. ( 1993 ) J. Mol. Biol. , 230 , 1260 – 1277 . Crossref Search ADS PubMed Kragelund , B.B. , Knudsen , J. , Poulsen , F.M. ( 1995 ) J. Mol. Biol. , 250 , 695 – 706 . Crossref Search ADS PubMed Krishnamani , V. , Hegde , B.G. , Langen , R. , Lanyi , J.K. ( 2012 ) Biochemistry , 51 , 1051 – 1060 . Crossref Search ADS PubMed Krishnamani , V. and Lanyi , J.K. ( 2012 ) Biochemistry , 51 , 1061 – 1069 . Crossref Search ADS PubMed Lu , D. , Liu , Z. , Wu , J. ( 2007 ) J. Chem. Phys. , 126 , 064906 . Crossref Search ADS PubMed Lundahl , P. , Greijer , E. , Sandberg , M. , Cardell , S. , Eriksson , K.-O. ( 1986 ) Biochim. Biophys. Acta , 873 , 20 – 26 . Crossref Search ADS Mandrup , S. , Højrup , P. , Kristiansen , K. , Knudsen , J. ( 1991 ) Biochem. J. , 276 , 817 – 823 . Crossref Search ADS PubMed Mattice , W.L. , Riser , J.M. , Clark , D.S. ( 1976 ) Biochemistry , 15 , 4264 – 4272 . Crossref Search ADS PubMed Micsonai , A. , Wien , F. , Kernya , L. , Lee , Y.H. , Goto , Y. , Refregiers , M. , Kardos , J. ( 2015 ) Proc. Natl. Acad. Sci. U. S. A. , 112 , E3095 – E3103 . Crossref Search ADS PubMed Montserret , R. , McLeish , M.J. , Böckmann , A. , Geourjon , C. , Penin , F. ( 2000 ) Biochemistry , 39 , 8362 – 8374 . Crossref Search ADS PubMed Oliver , R.C. , Lipfert , J. , Fox , D.A. , Lo , R.H. , Doniach , S. , Columbus , L. ( 2013 ) PLoS ONE , 8 , e62488 . Crossref Search ADS PubMed Otzen , D.E. ( 2011 ) Biochim. Biophys. Acta , 1814 , 562 – 591 . Crossref Search ADS PubMed Otzen , D.E. ( 2002 ) Biophys. J. , 83 , 2219 – 2230 . Crossref Search ADS PubMed Otzen , D.E. ( 2015 ) Curr Op Coll Interface Science , 20 , 161 – 169 . Crossref Search ADS Otzen , D.E. , Christiansen , L. , Schülein , M. ( 1999 ) Protein Sci. , 8 , 1878 – 1887 . Crossref Search ADS PubMed Otzen , D.E. and Oliveberg , M. ( 2002a ) J. Mol. Biol. , 315 , 1231 – 1240 . Crossref Search ADS Otzen , D.E. and Oliveberg , M. ( 2002b ) J. Mol. Biol. , 317 , 613 – 627 . Crossref Search ADS Paslawski , W. , Kristensen , J.V. , Lillelund , O. , Schafer , N. , Baker , R. , Urban , S. , Otzen , D.E. ( 2015 ) Proc. Natl. Acad. Sci. U. S. A. , 112 , 7978 – 7983 . Crossref Search ADS PubMed Pedersen , J.S. ( 2004 ) J. Appl. Crystallogr. , 37 , 369 – 380 . Crossref Search ADS Pires , J.M. , de Moura , A.F. , Freitas , L.C.G. ( 2012 ) Quim. Nova , 35 , 978 – 981 . Crossref Search ADS Pitt-Rivers , R. and Impiombato , F. ( 1968 ) Biochem. J. , 109 , 825 – 830 . PubMed Poghosyan , A.H. , Antonyan , L.A. , Arsenyan , L.H. , Shahinyan , A.A. ( 2014 ) Colloids and Surfaces A: Physicochemical and Engineering , 462 , 99 – 106 . Crossref Search ADS Poghosyan , A.A. , Arsenyan , L.H. , Shahinyan , A.A. , Koetz , J. ( 2016 ) Coll and Surf A: Phys Eng Asp , 506 , 402 – 408 . Crossref Search ADS Poghosyan , A.A. , Yeghiazaryan , G.A. , Gharabekyan , H.H. , Koetz , J. , Shahinyan , A.A. ( 2007 ) Mol. Simul. , 33 , 1155 – 1163 . Crossref Search ADS Rios-Lopez , M. , Mendez-Bermudez , J.G. , Dominguez , H. ( 2018 ) J. Phys. Chem. B , 122 , 4558 – 4565 . Crossref Search ADS PubMed Rizzuti , B. and Daggett , V. ( 2013 ) Arch. Biochem. Biophys. , 531 , 128 – 135 . Crossref Search ADS PubMed Sammalkorpi , M. , Karttunen , M. , Haataja , M. ( 2007 ) J. Phys. Chem. B , 111 , 11722 – 11733 . Crossref Search ADS PubMed Scheraga , H.A. , Khalili , M. , Liwo , A. ( 2007 ) Annu. Rev. Phys. Chem. , 58 , 57 – 83 . Crossref Search ADS PubMed Schwamberger , A. , De Roo , B. , Jacob , D. , Dillemans , L. , Bruegeman , L. , Seo , J.W. , Locquet , J.P. ( 2015 ) Nuclear Instruments and Methods in Physics Research Section B , 343 , 116 – 122 . Crossref Search ADS Shelley , J.C. and Shelley , M.Y. ( 2000 ) Curr Op Coll Int Sci , 5 , 101 – 110 . Crossref Search ADS Tanford , C. ( 1980 ) The Hydrophobic Effect. Formation of Micelles and Biological Membranes , 2nd edn. Wiley & Sons , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Vass , S. ( 1991 ) Struct. Chem. , 2 , (167)375–(189)397 . Vass , S. , Török , T. , Jakli , G. , Berecz , E. ( 1989 ) J. Phys. Chem. , 93 , 6553 – 6559 . Crossref Search ADS Veitshans , T. , Klimov , D. , Thirumalai , D. ( 1997 ) Fold. Des. , 2 , 1 – 22 . Crossref Search ADS PubMed Verlet , L. ( 1967 ) Phys. Rev. , 159 , 98 – 103 . Crossref Search ADS Walker , A.A. , Warden , A.C. , Trueman , H.E.W.S. , Sutherland , T.D. ( 2013 ) J. Mat. Chem. B , 1 , 3644 – 3651 . Crossref Search ADS Wu , C.S. , Ikeda , K. , Yang , J.T. ( 1981 ) Biochemistry , 20 , 566 – 570 . Crossref Search ADS PubMed Yonath , A. , Sielecki , A. , Moult , J. , Podjarny , A. , Traub , W. ( 1977 ) Biochemistry , 16 , 1413 – 1417 . Crossref Search ADS PubMed © The Author(s) 2019. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Molecular dynamics study of ACBP denaturation in alkyl sulfates demonstrates possible pathways of unfolding through fused surfactant clusters
JF - Protein Engineering Design and Selection
DO - 10.1093/protein/gzz037
DA - 2019-12-31
UR - https://www.deepdyve.com/lp/oxford-university-press/molecular-dynamics-study-of-acbp-denaturation-in-alkyl-sulfates-HT5kn7RBMg
SP - 175
VL - 32
IS - 4
DP - DeepDyve
ER -